**Theorising Indigenous Farmers' Utilisation of Climate Services: Lessons from the Oil-Rich Niger Delta**

### **Eromose Ehije Ebhuoma 1,\* , Mulala Danny Simatele <sup>2</sup> , Llewellyn Leonard <sup>1</sup> , Osadolor Obiahon Ebhuoma 3, Felix Kwabena Donkor <sup>1</sup> and Henry Bikwibili Tantoh <sup>1</sup>**


Received: 22 July 2020; Accepted: 4 September 2020; Published: 8 September 2020

**Abstract:** In the wake of a rapidly changing climate, climate services have enabled farmers in developing countries to make informed decisions, necessary for efficient food production. Climate services denote the timely production, translation, delivery and use of climate information to enhance decision-making. However, studies have failed to analyse the extent to which Indigenous farmers residing and producing their food in an environment degraded by multinational corporations (MNCs) utilise climate services. This study addresses this gap by analysing Indigenous farmers' utilisation of climate services in Igbide, Olomoro and Uzere communities, in the oil-rich Niger Delta region of Nigeria. Focus group discussions and semi-structured interviews were used to obtain primary data. Findings suggest that although the activities of Shell British petroleum, a MNC, have compromised food production, other factors have fuelled farmers' unwillingness to utilise climate services. These include their inability to access assets that can significantly scale up food production and lack of weather stations close to their communities needed to generate downscaled forecasts, amongst others. This paper argues that failure to address these issues may stifle the chances of actualising the first and second sustainable development goals (no poverty and zero hunger) by 2030 in the aforementioned communities.

**Keywords:** climate services; indigenous farmers; multinational corporations; systems thinking; Nigeria; sub-Saharan Africa

#### **1. Introduction**

Evidence in the literature suggests that climate variability and change have significantly compromised effective food production in sub-Saharan Africa (SSA) [1–3]. The radical transformation of a region once able to ensure that farmers could produce their food efficiently, to one where food productivity and output can no longer be guaranteed, due to climate variability and change, is partly responsible for the increased food and nutrition insecurity in SSA [4–6]. This has set in motion numerous unwanted responses in farming communities. These include climate-induced migration [3,7], engaging in more off-farm activities [8] and heavy reliance on government's food aid stamps [1], among others. These unwanted responses will likely be worsened, in part, by the aggravated difficulties expected to be associated with future food production due to increased occurrences of climate variability and change. Climate projections for SSA indicate that increased and unprecedented occurrences of extreme weathers will become the new normal by 2050 [9].

The growing concerns around SSA farmers' increased susceptibility to current and future climate variability and change have catalysed investments in climate services [2]. Climate services refer to the 'timely production, translation, delivery and use of climate data and information to enhance decision-making' [10] (p. 1). A huge wave of optimism exists within the scientific community about the capability of climate services to facilitate informed decisions, necessary to enhance food production among farmers [11,12]. Several studies underline the significance of climate services in scaling-up food production in the face of increased climate variability [2,13–15]. Yet, to our knowledge, no study has analysed the extent to which Indigenous farmers residing in communities with a history of fierce contestations with multinational corporations (MNCs) utilise climate services. By Indigenous, we mean 'those communities that claim a historical continuity with their traditional territories' [16] (p. 290). Indigenous people possess a peculiar culture and knowledge distinct to their community that have been tried and tested with real-life scenarios [17–19]. It is usually passed on from one generation to the next through oral communication and repetitive engagement [18].

MNCs, in their quest to exploit rich natural resources, have aggressively degraded the immediate environment that Indigenous farmers rely on to produce their food [20,21]. As highlighted in the Niger Delta region of Nigeria [22,23], Katanga in the Democratic Republic of Congo [24], Western Ghana [25] and Limpopo, South Africa [26], the activities of MNCs have adversely compromised farm yields and productivity. It is, therefore, crucial to ascertain if these cohorts of farmers trust and utilise climate services disseminated by government institutions. As Dutfield [19] (p. 24) argues, the exploitation of natural resources occurs through an 'unholy alliance of corporations with governments.'

Failure to deeply analyse the extent to which farmers, residing in regions where the activities of MNCs have degraded their natural environment, utilise climate services may undermine the actualisation of both the first and second sustainable development goals (SDGs) (no poverty and zero hunger) by 2030. This is especially if such farmers fail to utilise climate services to make informed farming decisions. Arguably, providers of climate services may be unaware of several underlying issues in communities affected by natural resources exploitation that may have been lingering for decades, which limits farmers' ability to produce food efficiently e.g., see [27–29]. This, in turn, may fuel scepticism among farmers regarding their respective government's agenda to invest in climate services over addressing the underlying issues that have negatively impacted food production, and consequently, undermine the use of such services.

Against this background, this paper investigates Indigenous farmers' perception of climate services. Also, it unpacks how the negative impacts of MNC activities influence Indigenous farmers' use of climate services. Further, it demystifies the attributes of those receptive to climate services. This paper analyses primary data obtained from the oil-rich Niger Delta region of Nigeria. The Niger Delta serves as an important case study especially since the region has been besieged by land degradation, for decades, due to crude oil exploration and exploitation by MNCs [19,22,30], which, in turn, has severely compromised food production [4,31]. Also, 'since 1970, oil revenues from Niger Delta have contributed over \$350 billion to Nigeria while the region remained one of the most impoverished parts of the country' [32] (p. 221). It is hoped that this paper will trigger discussions regarding factors that deter Indigenous farmers, whose livelihood activities and way of life have been adversely affected by the activities of MNCs, from utilising climate services. Such discussions are desperately needed to unravel effective interventions necessary to build the trust and confidence of Indigenous farmers in climate services.

This paper proceeds in five parts. The first provides a snapshot of existing literature regarding the role of climate services in upscaling food production. The second part succinctly illustrates how oil exploration activities by Shell British petroleum, a MNC, which results in oil spillages and gas flaring, have compromised food production and quality of crops harvested in the study areas, situated in the Niger Delta. It also highlights the frustrations of the people as they have not been beneficiaries of resources that can enable them produce food effectively as well as improve their quality of life. The third part provides a synopsis of the study areas and how primary data were obtained. The fourth part

brings to the fore the results. It aptly underlines the various factors responsible for the non-utilisation of climate services by Indigenous farmers. It also showcases the attributes of those willing to rely on climate services, if the factors undermining its use are decisively addressed. The final part discusses the implication of the findings.

#### **2. The Role of Climate Services in Scaling-Up Food Production**

A climate service is a decision aide that supports ex-ante climate risk management in agriculture, health, water, transport and other vulnerable sectors of the global economy [33]. The endorsement of the global framework for climate services, whose mission is 'to strengthen the production, availability, delivery and application of science-based climate prediction and services,' by delegates of 155 countries at the 2009 World Climate Conference III buttresses this point [34]. In SSA, the erratic and unprecedented occurrences of extreme weather conditions, especially in the last two decades, have crystallised the need for farmers to rely on climate services to make informed farming decisions [2,5,14,15]. Also, the drive to utilise climate services has been fuelled by the increased unreliability of Indigenous meteorological forecasts in some locations. e.g., see [35,36]. As Ouédraogo et al. [15] argue, climate services can assist farmers in reducing uncertainty and taking advantage of favourable weather information by planting intensively, and thus, minimise loss during unfavourable weather forecasts. In the case of extreme weather predictions, several pieces of literature have shown that climate services empower local and Indigenous farmers by allowing them to take measures aimed at protecting lives, livelihoods and household properties [37–39]. Climate services, therefore, support climate-resilient development.

In a simulation exercise conducted in Senegal, West Africa, Roudier et al. [2] found that farmers who used climate forecasts maximised benefits from predicted favourable conditions. In Burkina Faso, 'farmers exposed to climate information changed their farm practices based on the information they received, and that translated into better management of inputs to increase their farm productivity and improve their resilience to climate variability' [40] (p. 4). Somewhat similar findings have emerged from studies conducted in the Republic of Benin [41], India [42] and Uganda [14] respectively. Arguably, just when the future of food production seemed bleak for smallholder farmers due to the rapidly changing climate coupled with the increasing unreliability of Indigenous meteorological forecasts in some locations, climate services have rekindled farmers' hope to obtaining a meaningful livelihood. To substantiate this point, farmers in some SSA countries are now willing to pay for climate services [15,36]. This reinforces the World Meteorological Organization's [43] standpoint that the benefits of investments in climate services for agriculture and food security greatly outweigh the cost.

It is, however, noteworthy that the studies cited above were conducted in areas without substantial abundance of natural resources to trigger the influx of MNCs into such locations. If the theorisation of message interpretation by Sellnow et al. [44] is anything to go by, the chances of Indigenous farmers—who have and continue to be victims of the negative effects of the activities of MNCs—relying on climate services to make an informed decision will be slim. The reason for this, according to Sellnow et al. [44], is that for every message received, the audience draws conclusion by interpreting the conveyed message using availability heuristic, among others. Availability heuristic refers to 'a rule of thumb that allows people to solve problems based on what they remember and how easily their memory is retrieved, and how readily available that memory is' [45] (p. 54). The ease of retrievability is closely linked to significant landmarks in a person's life and could create signposts in one's memory, thereby making an experience easy to retrieve [46]. Thus, availability heuristic, which influences trust and credibility [47,48], may negatively impact the adoption and utilisation of climate services by Indigenous farmers whose livelihoods have been compromised by the activities of MNCs.

#### **3. The Delta State: Background Information**

The Delta state comprises one of the nine states in Nigeria that make up the oil-rich Niger Delta region. Crude oil—the mainstay of the nation's foreign reserves and GDP—is mainly obtained from the Niger-Delta [23,27], a region which constitutes approximately 8% of the nation's total landmass (Figure 1). Anecdotal evidence suggests that the second crude oil discovery happened in Uzere in 1958, after the first discovery was made in Oloibiri, Bayelsa state, in 1956 [49]. Igbide, Uzere and Olomoro communities have approximately 62 oil wells between them. The aforementioned communities, as well as the entire Niger Delta region, have contributed immensely to Nigeria's gross domestic product (GDP) due to no less than four decades of consistent oil exploration and exploitation activities, amounting to 90, 000 barrels of oil daily [50,51]. However, these communities have gained a global reputation for the negative impacts that crude oil exploration has had on the livelihoods and general welfare of its inhabitants [23]. Specifically, oil spillages and gas flaring, which have adversely compromised the health of local people, their ability to produce food effectively and the nutritional quality of crops harvested, have consistently made headlines. Consequently, this has metamorphosed a relatively tranquil region, in the 1960s, to one prone to conflicts with MNCs, since the 1990s.

**Figure 1.** Map of the Niger-Delta region in Southern Nigeria.

No fewer than 30 protests and demonstrations against MNCs (some relatively peaceful and others extremely violent) as well as intermittent disruption of oil exploration in all three communities have been recorded since the 1990s [32,51,52]. According to the president-general of Isoko Development Monitoring Group (IDMG):

'Oil exploration has literally killed the fruitfulness of Isokoland. Our farmers are crying, lamenting their ordeal in the hands of the oil companies in their lands that have refused to pay them meaningful compensations' [53] (p. 1).

Compounding the woes of farmers is their inability to easily access fundamental assets that can significantly improve their welfare. This is a factor at play due to corrupt practices among government officials. Fundamental assets refer to the financial, natural, social, human and natural assets or capital (Table 1). As a result, these communities, including their community members, continue to live in abject poverty without basic infrastructures such as good roads, potable water and constant power supply [23]. Hence, these communities and the entire Niger Delta are classified as a paradox; oil-rich but impoverished [30].


**Table 1.** Definition of the fundamental bundle of assets or asset portfolio.

Sources: Bebbington [54]; Thornton et al. [55]; Moser and Satterthwaite [56]; Moser [57].

#### **4. Materials and Methods**

#### *4.1. Study Area*

Igbide, Uzere and Olomoro communities are located in Isoko south local government area (ISLGA) in the Delta state of Nigeria (Figure 2). *Isoko* is the local dialect spoken in these areas. The region has a mean annual precipitation of 2500–3000 mm [4]. The weather patterns can be categorised into rainy and dry seasons [58]. Normally, the rainy season commences fully in June and lasts until October. This is closely followed by the dry season which commences in November until early March. In the dry season, dry and dusty conditions known as *harmattan* occur between mid-December and late January.

Small-scale farming is the primary economic driver in these three communities. Cassava (*manihot esculenta*) and groundnuts (*arachis hypogea*) are the major crops produced annually together with okra, pepper, sweet potatoes, yam and plantain. Cassava constitutes 60% of calories consumed. With the exception of cassava, the crops produced depend predominantly on the early rains. Groundnuts, however, are highly sensitive to rising temperatures. Some elderly farmers (above 50 years old) lamented about how the scorching sun, between February and April, has significantly reduced the crop's output by nearly 50% in comparison to the bountiful harvests they were used to in the 1960s. Presently, when they harvest, some of the groundnuts are empty pods. The farmers' explained that reduced groundnut output, which has become the norm, started from the 1990s.

The farmlands in Igbide and Uzere are low-lying, while Olomoro comprises both low and high-lying farmlands. The low-lying farmlands in each of the study areas experience seasonal flooding from mid-June, at the earliest, to the last week in October every year. In extreme conditions, the low-lying farmlands remain inundated until the third week in November. Thus, cassava, which requires a minimum of six months to attain maturity, is usually planted in December and harvested between June and August each year on the low-lying farmlands.

**Figure 2.** Map of the study areas.

#### *4.2. Methodology*

Qualitative methods were adopted to offer deep insights usually hidden from statistical analysis. The study on which this article is based, formed part of a larger undertaking comprising 35 focus group discussions (FGDs) and 14 one-on-one semi-structured interviews, conducted between June and October 2015. Follow up interviews were conducted in July 2016. These discussions explored the participants' perceptions of climate-related threats to food production, how they adapt to such threats, and the extent to which they use government-issued seasonal forecasts, among other issues. Each focus group comprised between three and twelve participants. Two-thirds of the participants were between the ages of 42 and 85 (median = 64) and had no formal education. This article draws on 22 FGDs and five one-on-one, semi-structured interviews where participants provided valuable insights regarding their perception of climate-related threats to food production, how they adapt to such threats, and the extent to which climate services influence their farming decisions, among other issues. Key questions asked include *do you have any local sign(s) that you rely on to know when the early rainfall or rainy season is about to start so that you can start preparations for production or the best time to*

*start planting? How accurate has the local sign(s) been in predicting the weather? Have you ever relied on weather information from NIMET when preparing and planning for a planting season*? *Did you receive the 2012 flood warning by NIMET? Did you receive the 2013 flood warning predicted by NIMET?* If yes, *how did it influence your farming decisions?* And *how willing are you to rely on scientific weather predictions in the future?* Some FGDs and semi-structured interviews were conducted in *Isoko* and translated to *Pidgin* English by field assistants to enable the first author to understand their responses. Other interviews were conducted in *Pidgin* English. Both the FGDs and interviews were edited to formal English during transcription of the audio recordings and have been presented as edited.

Eligible participants were selected based on the following criteria: the individuals had to have been farming in one of the study areas for a minimum of ten years, gender (both had to be represented), their household assets and livelihoods had to have been adversely affected by the 2012 flood disaster, those who produced most of their food on low-lying farmlands, and their willingness to participate in the study. The low-lying farmlands are usually inundated around June to October annually. The qualitative data obtained were analysed using thematic analysis.

#### **5. Results**

#### *5.1. Perception of Climate Services*

When participants were asked if they relied on climate services to make farming decisions, they explicitly highlighted that they do not utilise the information. When probed as to why they do not utilise such information, most revealed that they rely on their local and indigenous knowledge systems (LIKS). These include in-depth historical knowledge of the weather patterns, croaking of frogs and appearance of red-like millipedes. Others include the flowering of rubber trees (*ficus elastic decora*), greening of cassava leaves, lunar observation in December, and the height at which the weaver bird constructs its nest on a tree [59,60]. These Indigenous meteorological indicators help to forecast the commencement of the rainy season, as well as the anticipated total rainfall in a farming season. While the participants acknowledged that their LIKS have not always been entirely accurate, the majority, however, revealed that they are not keen to switch to utilising climate services for the following reasons:

#### 5.1.1. 'Suffering in the Midst of Plenty' Syndrome

For the majority, it defied human comprehension for the Nigerian government to invest in climate services when households continue to live in abject poverty in the midst of plenty. Despite consistent exploitation of crude oil for over four decades, their communities have remained inconceivably underdeveloped due to a dearth of communal physical assets. Because they practice annual cropping, harvesting of cassava coincides with the rainy season. Thus, it is exasperating and expensive for households to transport their produce to neighbouring communities to sell as transport drivers are forced to take alternative routes that take much longer. This is due to the numerous potholes on the roads that are tantamount to death traps during the rainy season when they are filled with rainwater. Also, constant power supply is non-existent in ISLGA. This has resulted in the few privileged farmers—who own refrigerators—not being able to cultivate perishable crops like okra and pepper in large quantities, as they cannot prevent most of the produce from rotting away. Furthermore, farmers' inability to access financial assets is a major deterring factor in their struggle to live above the global poverty line of USD 2 daily.

While participants' acknowledged that the government occasionally made provisions to furnish farmers with farm loans and other incentives, such as fertilizers and farm machinery, the majority stated that they have never been beneficiaries. Some argued that they only became aware of such opportunities after the disbursement process was over. Others revealed that the loan distribution process is marred by corruption and that the key beneficiaries were 'ghost' farmers and relatives of those in charge of disbursing the loans. In venting his frustration, one participant between the ages of 40 and 50 lamented:

'Government is not helping anybody from this area despite being an oil producing community ... We only hear it on the radio of the various incentives given to assist farmers, but it is not getting to those on the grassroots. Instead, the beneficiaries are those in the hierarchy such as the executives of small-scale farmers' group association who use it for their private farms.'

It is important to note that some of the participants skilled in generating Indigenous weather forecasts attributed the increased anomalies of their LIKS to gas flaring in Oleh, a community approximately 20 km away from the study areas. An elderly participant, aged between 60 and 70 years, revealed:

'Gas exploration and flaring have seriously affected the trees we use to predict the weather because they do not grow well as they are supposed to. It becomes difficult for our predictions to be correct.'

Also, another elderly participant aged between 70 and 80 years asserted that gas flaring impairs her vision, which is critical to predicting if it would rain or not, through cloud observation. However, an elderly participant also aged between 70 and 80 years argued that gas flaring could not be the reason for the inconsistencies of their LIKS, but admitted that it kills their crops and damages the land. She further pointed out that the quality of *garri* produced has declined because the soil nutrients have been compromised due to oil exploitation activities, a viewpoint shared all the study participants. A few participants, however, added that farmers' inability to practice bush fallowing, due to sporadic increase in population, is also a contributing factor. In this regard, an elderly participant between the ages of 60 and 70 years argued:

'Before Shell commenced oil exploration activities, we were told that at a particular stage, they would fertilise the whole land ... They said they would inject the land for them so that it will replenish the soil. The aim was to ensure that the soil will always be fertile for agricultural production ... But up till date, this has not been carried out.'

Also, focus group participants in Igbide revealed that gas flaring expedites the corrosion and decay of zinc roofing sheets—the most affordable materials—used by households to roof their houses. A participant between the ages of 50 and 60 years categorically stated:

'If you roof your house with iron zinc, within 4 to 6 years, it will start decaying. When it starts raining, the roofs start leaking, and then you have no option but to change the roof. This is causing people economic setback because the money they are supposed to use for other things are now used to repair or replace their damaged roofs. If you don't have money to buy the aluminium Cameroon zinc, before you leave this world you will probably have to roof your house for about seven times.'

The adverse effects of oil exploration activities on food production (Figure 3) coupled with the history of neglect by various government regimes inter alia made most of the participants question the rationale behind investing in climate services. They cannot reconcile how their region can be endowed with huge reservoirs of crude oil and yet the indigenes continue to live in abject poverty. From the participants' perspective, providing and ensuring easy accessibility to assets should be the first step to enhancing food production for farmers because they understand the weather patterns in their community. In this regard, a participant between the ages of 40 and 50 years stated:

'It is only a foolish person that does not understand the terrain. When you plant cassava on 10 plots, like me, as from next month [referring to July] I will start to harvest my cassava. I will not wait till August when the flood usually inundates the farmlands, that is how we cope. A wise person will remove his/her cassava from the farmland before that time.'

**Figure 3.** Causal loop diagram illustrating how oil exploration compromises food production (B = balancing feedback loop). N.B. If an increase in one variable leads to an increase in another variable, it is denoted by '+' close to the arrow head, while if it results in a decrease in another variable, it is denoted by '−' close to the arrow head. The double parallel red lines indicate a delay between cause and effect.

#### 5.1.2. Doubts Surrounding the Accuracy of Meteorological Forecasts

The credibility of meteorologists in accurately predicting useful weather information was severely questioned in the aftermath of the 2013 floods that never made landfall in the Delta state as forecasted by the Nigerian meteorological agency (NIMET). In the seasonal climate outlook for 2013, NIMET stated that 31 states were likely to be hit with the same magnitude of flood that occurred in 2012. The 2012 floods wreaked devastating havoc in 31 states of the federation, that was accurately predicted [61,62]. The 2013 flood forecast pinpointed the Delta state as one of the hotspots where the worst impact of the looming disaster would occur. The forecast spread like wildfire in Igbide, Uzere and Olomoro communities, where the 2012 flood wreaked devastating havoc partly because most participants did not receive any warning due to the government's overwhelming reliance on radio and television broadcasts, which was a gross mismatch for the farmers.

This time around, participants became aware of the 2013 forecast through word of mouth from relatives living in urban areas as well as fellow villagers. Consequently, the majority embarked on the inconvenient but inevitable task of adopting proactive measures as they were determined not to be caught unawares again, as they were in 2012. Thus, most participants completely abandoned their farmlands in the low-lying regions for fear of their produce being destroyed by floodwaters. They also sought high-lying farmlands in locations both within and outside their communities that were unaffected by the 2012 floods. The quest to hire or lease viable farmlands in high-lying areas resulted in the sporadic hike in rental price. A few participants, however, produced food on their low-lying farmlands, but in negligible quantities and focused more on crops that matured early. As one participant between the ages of 50 and 60 years explained:

'Farmers who used to cultivate in farmlands close to Urie Lake [farmlands in this region are inundated to about 2–2.5 m between August and October annually] were afraid of planting when they heard that another flood was coming. They left Igbide to neighbouring communities that were unaffected by the 2012 floods looking for farmlands to grow their food. Some were already contemplating relocating from the community. Others planted on their low-lying farmlands but in small quantities. The sudden panic amongst farmers made it extremely costly to plant during that period.'

However, the participants' decision to adopt several pro-active measures was met with unpleasant consequences (Table 2), which had severe implications for their livelihoods as the flood did not occur, at least in the Delta state. To compound their woes, farmers were neither provided with an explanation why the forecast was imprecise nor given relief measures to minimise the hardship brought about by 'inaccurate forecast-induced food insecurity.' In narrating the impact that the wrong forecast had on household food security, a participant between the ages of 50 and 60 years stated:


**Table 2.** Proactive measures adopted by farmers in response to the seasonal climate forecast in 2013.

'As a result of the news, people refused to go and plant. This resulted in severe hunger for a lot of people in the community. Me, myself, I did not plant because of the news.'

In the same vein, another participant between the ages of 40 and 50 years explained:

'We use to supply *garri* (processed cassava) to Bayelsa and Edo states prior to that 2013 forecast. Because we adhered to the rumour, we were now the ones going to those states to buy garri. The painful part was that the cost to purchase the produce was extremely expensive. So this time around, if the government distributes any forecast that says there will be extreme flooding, I will ignore the forecast and plant normally.'

The experience has instilled in the minds of the people that their welfare is of little or no concern to the government, and consequently, has eroded their trust in climate services.

#### 5.1.3. Mediums of Disseminating Weather Information

The channel of communication used to disseminate weather and climate information to households in the Delta state may be a factor undermining its use to make informed farming decisions. Radio and television broadcasts are the mediums used to convey weather information to households. This may arguably be the reason why nearly all the participants, with the exception of one person, did not receive the 2012 flood warning, although it is not entirely clear if they would have adopted pro-active behaviours as their LIKS did not predict that such event would occur. With the exception of Sundays when they attend church services and social gatherings as well as market days—once in four days—when they sell their crops, the farmers' typical daily routine entails leaving their homes before 7 a.m. for their respective farmlands. This is to ensure that as much work as possible is done before the scorching noonday and afternoon sun emerges. Thereafter, they seek shade to get a reprieve from the heat before continuing their work. Consequently, the farmers usually return home late in the afternoons or evenings extremely fatigued, and thus, barely have any iota of strength left in them to listen to both radio and television broadcasts.

Also, the lack of constant power supply is another factor that compromises the effectiveness of radio and television broadcasts in the region. In lamenting on the despicable state of power supply in the region, a participant aged between 60 and 70 years argued:

'We are just paying for ordinary power cables that do not provide us with electricity. For more than four days now, there has been no electricity.'

In terms of the most effective channels for communicating important information to households, the participants unequivocally argued that the use of community town-criers and town-hall meetings remain the most effective mediums of communication in their respective communities. In addition, a participant from Uzere, in his early 80s, recommended that local people should be trained on how to interpret weather information. This is because:

'NIMET and agricultural extension workers will not have the time to go round the nooks and cranny of the community to inform people of what the weather outlook for the forthcoming farming season will be like, as against their people that reside in the community.'

#### 5.1.4. Absence of Weather Stations

At the time of data collection, the closest weather station to the study areas was located in Warri, a city approximately 60 km away from the study sites. The other station was in Asaba, about 140 km away from the study sites. The implication is that the forecasts that were generated were not tailored to ISLGA. It is, therefore, not inconceivable to argue that the lack of downscaled forecasts may have deterred some farmers from listening to both radio and television broadcasts for weather-related information. Also, this may partly explain why the participant (a woman from Igbide, in her 50s) who heard the flood warning over the radio failed to act proactively. She remarked:

'Before the flood occurred, it was announced on the radio that there was going to be a flood incident, which would affect most parts of the nation. However, due to a lack of scientific understanding of the message's content and not being certain that Igbide will be affected, I ignored the warning.'

It is, however, noteworthy that a small minority was open to the idea of using climate services to inform farming decisions.

#### *5.2. Characteristics of Participants Willing to Utilise Climate Services*

Two groups of participants were open to using climate services to make farming decisions. The first group open to the idea of utilising climate services were young male participants (under 40 years) who had obtained at least a high school education. The young male participants assisted their parents in their respective fields while being actively involved in off-farm activities, such as electronics repairs, carpentry and welding. With regards to the perceived role climate services can play in facilitating effective food production, a participant, in his early 30s, commented:

'If farmers in this community can be getting the seasonal rainfall predictions, it will be extremely helpful in protecting our crops and other livelihood assets ... serious farmers will need this sort of information.'

The next group of farmers that were willing to utilise climate services to make informed farming decisions were the elderly people (above 50 years old) that were educated and had a viable alternative source of livelihood. Nonetheless, an individual that fell within this bracket sounded disinterested. The elderly participants interested in climate services were retired civil servants who received a regular monthly pension. They had properties such as vast hectares of farmlands, houses and fishing boats they leased out to earn regular income. In this regard, a pensioner from Uzere aged between 60 and 70 years, who had obtained a university degree, expressed his willingness to utilise climate information by stating that:

'If farmers can get the meteorological information before the planting season (November), it will play a significant role in protecting the assets of the small-scale farmers. If I can get the weather information for this community, I will take full advantage of it.'

#### **6. Discussion**

In the wake of a rapidly changing climate, local farmers' utilisation of climate services in SSA is crucial to facilitating informed farming decisions necessary to scale-up food production [2,14,41,42]. In contrast to their counterparts in various SSA countries that have embraced this technology, local and indigenous farmers in the Delta state are yet to adopt climate services, with the majority not keen to do so. The indigenous farmers are disillusioned that despite their communities' significant contribution to Nigeria's GDP, basic amenities and infrastructures—consistent power supply, potable water and good roads—are virtually non-existent. Also, despite the farmers' overwhelming dependence on food production to obtain a livelihood, the majority have never been beneficiaries of vital assets such as farm loans and machinery. This is against the backdrop of the Delta state government and federal government making such schemes readily available [23,63,64]. Thus, the notion that climate services—an 'abstract entity'—can scale-up food production seems utopic. From the farmers' viewpoint, their community symbolises an express tunnel for capital accumulation, with MNCs and government being the sole beneficiaries. Also, the fact that their LIKS used in predicting weather conditions have been negatively affected by oil exploration and exploitation activities, as well as farmers being the victims of a wrong forecast cast further doubt on the efficacy of climate services to upscale food production.

Studies conducted in various developing countries that acknowledge indigenous farmers' willingness to rely on [33,65] and actual utilisation [14,15,36] of climate services, the environment they depend upon to generate indigenous meteorological forecasts have not been exploited by MNCs. As studies in Thailand [66], Ethiopia [67] and Nigeria [68] show, a catalyst to a decline and or loss of LIKS is anthropogenically-induced modification of the environment due to natural resources exploitation. Nonetheless, the indigenous farmers in the Delta state may change their current standpoint of not being willing to utilise climate services if access to fundamental assets that can significantly improve their welfare are made readily available. This hinges on the fact that, despite lack of trust in the

government coupled with inability to access capital assets, they still relied on the 2013 forecast to make farming decisions.

As documented by Singh et al. [46] (p. 2428), people's memories give 'weightage to recent events because the consequences are perceived more strongly than those in the distant past'. Thus, easy access to assets, coupled with situating weather stations in ISLGA that are 30 km2 apart (recommended guideline to generate accurate local downscaled forecasts [69]), may cause farmers to recalibrate the incorrect 2013 forecast experience signpost in their memory as a one-time event with a low probability of recurrence. To advance the adoption of climate services, it might be necessary to work closely with those willing to utilise climate services, perhaps through pilot studies as a useful starting point. Showcasing the effectiveness of climate services in upscaling food production versus relying solely on LIKS might be the propelling force that will get other farmers to change their standpoint. Such pilot studies will address Serra and McKune's [70] (p. 4) concern that, 'while much attention has focused on improving coverage of climate services, we should equally invest resources and efforts in tackling the complex domain of individual and collective perception of information, and consequent behavioural change.'

If and after succeeding in convincing farmers on the effectiveness of climate services, and in the eventuality of another wrong future forecast, although rare if downscaled forecasts are produced but not entirely inevitable, see [71], government must realise that the adverse impacts of such occurrence on farmers' livelihood almost mirrors the consequences of an extreme weather disaster. Therefore, such occurrences should be handled in similar magnitude as an extreme weather event, which often triggers government's immediate response in mobilising resources to assist victims. If farmers are convinced that there will be a safety net in the aftermath of an erroneous forecast, the likelihood that they will utilise climate services may be high.

However, caution should be exercised when interpreting the findings of this study because the farmers' advocacy for assets over climate services may be attributed to the fact that only groundnut production has been adversely affected by climate variability and change. Thus, they may have become accustomed to the shock, especially since the reduction in groundnut production has occurred for over three decades. The regular occurrences of low groundnut outputs, which have reduced the monetary income received from its sale, may have desensitized farmers to the shock. This experience is in stark contrast to cassava—the main staple consumed by farmers—whose production has not been aggressively impacted by climate variability and change. The major deterrent to maximising cassava production is the annual seasonal flooding (mid-June to last week in October) of their low-lying farmlands, which is not perceived as a challenge because it has been the norm in the respective communities [4]. This may provide a valuable justification why they are not keen to utilise climate services, unlike farmers in India and Bangladesh [39] where the major staples consumed are highly sensitive to climate variability and change. As a result, these farmers rely on the technology to improve farming decisions. Nonetheless, this study offers valuable insights into factors that may undermine the use of climate services in areas that have been exploited by MNCs.

#### **7. Conclusions**

In the face of increasingly palpable impacts of climate variability and change amongst farmers in developing nations, climate services are regarded as essential to such farmers making informed decision for efficient food production. Moreover, climate services have recorded tremendous success to the point that farmers are now willing to pay for such services. Despite this development, studies have failed to understand the extent to which Indigenous farmers residing in communities with a history of fierce contestations with MNCs utilise climate services to increase agricultural productivity. It has been observed that MNCs, in their quest to exploit rich natural resources, aggressively degrade the immediate environment that Indigenous farmers rely on for food production. Since the exploitation of natural resources by MNCs occur through an alliance with a country's government, it is necessary to ascertain how the negative impacts of MNC activities influence Indigenous farmers' perception and utilisation of climate services.

The findings of this study suggest that indigenous farmers in the research sites where data for this study was collected do not rely on climate services but on their LIKS such as the croaking of frogs, and lunar observation in December amongst others. Although these farmers do acknowledge that their LIKS have not always been entirely accurate, the majority suggested that they were not keen on utilising climate services. A reason for this standpoint is because the participants continue to be victims of what we refer to as 'suffering in the midst of plenty' syndrome. They cannot reconcile how people continue to live in abject poverty, and their communities have remained underdeveloped despite the significant contributions crude oil explored from their locality make to Nigeria's GDP. While the farmers acknowledge that the government occasionally made provisions to provide farmers with farm loans and other assets like farm machinery, most have never been beneficiaries. Also, participants asserted that the quality of *garri* produced has declined because the soil nutrients have been compromised due to oil exploitation activities. Thus, the adverse effects of oil exploration activities on food production and the influence it has on the accuracy of their LIKS (Figure 3) coupled with the history of neglect by various government regimes inter alia made most participants question the rationale behind investing in climate services.

Furthermore, other factors beyond the activities of MNCs have fuelled farmers' unwillingness to utilise climate services. These include non-usage of their local channels of communication—town-hall meetings and town-criers—to disseminate weather information, lack of weather station close to the communities to generate accurate downscaled forecasts and being victims of an erroneous scientific forecast in 2013. Notwithstanding, the study finds that young males (under 40 years) and a few older adults (over 50 years old) are willing to utilise climate services if the factors undermining the production of accurate forecasts are addressed. The commonalities among these cohorts of farmers are that they have obtained at least a high school education and have viable alternative sources of livelihood outside of farming. Given this scenario, addressing the challenges that hamper farmers' utilisation of climate services in Igbide, Olomoro and Uzere communities is now a matter of urgency especially if the Nigerian government is keen on actualising the first and second SDG goals by 2030. Failure to address these issues will be a massive blow for the incumbent government that has been investing heavily in agriculture, deemed a viable alternative to successfully diversify Nigeria's economy from its extensive reliance on crude oil for most of its foreign exchange. Besides erecting weather stations in ISLGA that are 30 km<sup>2</sup> apart, a useful starting point will be to conduct pilot studies with those willing to utilise climate services. Highlighting the effectiveness of climate services in upscaling food production versus relying solely on LIKS might be a catalyst that will convince other farmers to use climate services in making informed farming decisions.

**Author Contributions:** Conceptualization, visualization, design of research instruments, E.E.E.; methodology, E.E.E. and O.O.E.; data collection, writing—original draft preparation, E.E.E.; writing—review and editing, F.K.D. and H.B.T.; supervision, M.D.S. and L.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** The research funding provided by the National Institute for the Humanities and Social Sciences and the Social Sciences Council for the Development of Social Science Research in Africa (NIHSS-CODESRIA) is gratefully acknowledged. The first author acknowledges the South African System Analysis Centre (SASAC) and the funding provided by the National Research Foundation (NRF) and the Department of Science and Technology (DST) of South Africa.

**Acknowledgments:** The generous comments of all three reviewers are deeply appreciated.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Groundwater Resources in the Main Ethiopian Rift Valley: An Overview for a Sustainable Development**

**Sabrina Maria Rita Bonetto , Chiara Caselle \* , Domenico Antonio de Luca and Manuela Lasagna**

Department of Earth Science, University of Turin, 10125 Turin, Italy; sabrina.bonetto@unito.it (S.M.R.B.); domenico.deluca@unito.it (D.A.d.L.); manuela.lasagna@unito.it (M.L.)

**\*** Correspondence: chiara.caselle@unito.it

**Abstract:** In arid and semi-arid areas, human health and economic development depend on water availability, which can be greatly compromised by droughts. In some cases, the presence of natural contaminants may additionally reduce the availability of good quality water. This research analyzed the water resources and hydrochemical characteristics in a rural area of the central Main Ethiopian Rift Valley, particularly in the districts of Shashemene, Arsi Negelle, and Siraro. The study was developed using a census of the main water points (springs and wells) in the area and the sampling and physico-chemical analysis of the water, with particular regard to the fluoride concentration. In many cases, fluoride content exceeded the drinking water limits set by the World Health Organization, even in the absence of anthropogenic contamination. Two different aquifers were recognized: A shallow aquifer related to the eastern escarpment and highlands, and a deep aquifer in the lowland areas of the rift valley on the basis of compositional changes from Ca–Mg/HCO3 to Na–HCO3. The distribution of fluoride, as well as pH and EC values, showed a decrease from the center of the lowlands to the eastern highlands, with similar values closely aligned along an NNE/SSW trend. All these data contribute to creating awareness among and sharing information on the risks with rural communities and local governments to support the adequate use of the available water resources and to plan appropriate interventions to increase access to fresh water, aimed at the sustainable human and rural local development of the region.

**Keywords:** groundwater resources; fluoride; main Ethiopian Rift Valley

#### **1. Introduction**

The regional planning and the management of natural resources require considering the interactions among human needs, ecosystem dynamics, and resource sustainability. Human needs and economic activities, such as industry, agriculture, and animal husbandry, require a continuous water supply. Over the past decades, water use has more than doubled [1], and the water demand will further increase due to a growing global human population. However, both the availability and the quality of water resources will be affected by socio-economic and technological developments, climate change and increasing climate extremes, such as droughts and floods, particularly in developing countries [2,3]. Demographic growth and unsustainable economic practices are affecting the water quantity and quality, making water an increasingly scarce and expensive resource, especially for the poor, the marginalized, and the vulnerable [4]. These factors will make it difficult to achieve the Sustainable Development Goal SDG 6 (one of the United Nations' Sustainable Development Goals (SDGs) for the year 2030), which requires sustainable management of clean accessible water for all. To achieve Goal 6, broad and in-depth knowledge of the global dynamics of water use and availability is necessary. Sustainable management of water resources for different uses will not only need to account for demand in water quantity, but also for water temperature, nutrient levels, and other pollutants [5].

The availability and quality of fresh water are particularly important in arid and hyper-arid environments, where groundwater plays an important role in supporting the

**Citation:** Bonetto, S.M.R.; Caselle, C.; de Luca, D.A.; Lasagna, M. Groundwater Resources in the Main Ethiopian Rift Valley: An Overview for a Sustainable Development. *Sustainability* **2021**, *13*, 1347. https://doi.org/10.3390/su13031347

Academic Editor: Maurizio Tiepolo Received: 20 December 2020 Accepted: 25 January 2021 Published: 28 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

economy, being the main source of water [6,7]. In these environments, the geological setting commonly contains bedrock of crystalline rocks characterized by secondary porosity due to the presence of interconnected fractures, faults, and shear zones, which represent a favorable setting for hosting and channeling groundwater [8,9]. Therefore, groundwater flow and yield in these aquifers are strictly related to the presence of lineaments and structural features in the basement [10–12]. An additional supply of water is represented by loose deposits, such as fluvial or lacustrine deposits, that are randomly distributed on the surface [13].

The success of expensive drilling campaigns is influenced by the complexity of the geological, structural, and hydrogeological setting of many arid environments and by the lack of subsoil investigations [14]. More specifically, geological features greatly affect the yield and quality of groundwater resources. As a consequence, high concentrations of many chemical elements can occur naturally in groundwater due to the local geology and to water interactions with rocks. Several studies [15–19] have shown that water in areas with particular geological features did not meet established drinking water limits for the presence of natural geogenic contaminants without influence from anthropogenic causes.

This study focuses on the characteristics of groundwater quality in the Oromia Region of the south-center portion of the Main Ethiopian Rift (MER), which represents the northern sector of the East African Rift System. In this region, rocks are mostly volcanic, and volcanic activity is still ongoing. Only a few perennial streams are present at the surface and, despite the presence of many lakes, the water supply in the area mainly depends on rainfall because of the poor physical and chemical quality of surficial water. More specifically, due to the scarce data about water availability, the paper is mainly focused on groundwater quality in the study area.

Natural sources of elevated fluoride concentrations in groundwater and lakes in the MER have been reported in the scientific literature [20–23] and, therefore, water is not always suitable for human consumption. Indeed, prolonged ingestion of drinking water containing F− ions that exceed the tolerance limits can cause dental and skeletal fluorosis, with negative effects on children and young people [24].

The concentration and mobilization of F− ions and other elements in groundwater are dependent on the chemical and physical processes that occur between the water and the geological environment. The occurrence of F− in water is generally associated with volcanic rocks, particularly in the high-temperature geothermal settings that are common at convergent plate boundaries, as well as in intraplate areas characterized by extensional tectonic activity, such as the African Rift [22].

The World Health Organization recommends a maximum concentration limit of 1.5 mg/L, in the case of hot water (above 25 ◦C), or in tropical countries where the daily intake of drinking water is high, the value decreases to 0.7 mg/L [25]. High concentrations of fluoride affect human health. Fluoride may give rise to mild dental fluorosis at concentrations between 0.9 and 1.2 mg/L in drinking water [26]. This is particularly true in warmer areas, where dental fluorosis occurs with lower concentrations in drinking water because of the greater amount of water consumed [27,28]. Fluoride can also have more serious effects on skeletal tissues (bones and teeth) with long-term ingestion, particularly if the drinking water contains 3–6 mg of fluoride per liter [28].

In reason of the reported presence of fluoride contaminations in surface and groundwater in neighboring areas, the present study aims to propose a census of wells and springs in the Oromia Region and qualitative analysis of the water resource available for the local communities, with particular attention to the presence and distribution of fluorides. The outcomes of the study may provide important elements for a thoughtful decision-making process aimed at the sustainable human and rural local development of the region. We focused, more specifically, on a countryside area of the West Arsi Zone (Oromia Region) in the rural district of Arsi Negelle, Shashamane, and Siraro. The area is characterized by flat land, and the economy is based upon the integration of smallholding agriculture and livestock. As a consequence, life in the community and rural activities are tightly connected to the water supply [29–32]. This rural area is very populated, but only 30% of the people have access to water, and rural communities are at greater risk from meteorological, hydrological, and agricultural drought [33,34]. The results of the geographical distribution and physico-chemical features of the water resource were interpreted and discussed on the basis of the geological setting of the area, creating a global picture of the water availability and quality.

#### **2. Geological Setting**

The study area is located in the central sector of the Main Ethiopian Rift Valley (MER). The MER is a NNE-SSW trending segment of the Rift Valley, and it is bordered by the Ethiopian Plateau to the west and the Somali Plateau to the east by means of evident fault slopes belonging to the main Rift Valley System Fault [35].

Lowlands (approximately 1600 m a.s.l.), transitional escarpments and highlands (approximately 2500 m a.s.l.) are the main geomorphologic environments of the MER. Many volcanic lakes are present in the lowlands. The lakes have had a recent evolution marked by transgressive and regressive phases, each phase followed by deposition of fluvio-lacustrine deposits.

The bedrock in the highlands consists predominantly of basic volcanic rocks (lava and ash, mainly of Tertiary age), whereas the bedrock in the lowlands consists mainly of acidic volcanic rocks (peralkaline silica-rich rhyolitic ignimbrites, including ash and pumice). Weathered and reworked volcanic rocks, silicic tephra, and small alluvial fans occur along the escarpments and the border of the highlands (Figure 1). Eluvial lateritic crusts are also present and consist of clay, silt, and fine sand [20,21].

The MER is geographically divided into three sectors: (i) NMER (northern MER), from the Afar depression to near Lake Koka; (ii) CMER (central MER), from Lake Koka to Lake Awasa through the lake region; and (iii) SMER (southern MER), from Lake Awasa to the broadly rifted zone of southern Ethiopia [36] (Figure 1).

The part of the Oromia Region is included in the CMER and is located in the Zway-Shala basin area. This sector is characterized by bedrock mainly consisting of volcanic products (pyroclastic rocks, rhyolites, tuffs, and basaltic lava flows), and volcano-lacustrine and fluvio-lacustrine deposits (clay, silt, sand, and gravel interbedded layers) [35,37–39]. In particular, the study area contains the Langano, Abijata, Awasa, and Shala Lakes, which are the remnants of a single wide ancient lake that extended to the area where upper Quaternary fluvio-lacustrine deposits are now observed [35]. Some lakes are connected by an ephemeral drainage system consisting of a few perennial streams and are closely linked with the groundwater system [40].

**Figure 1.** (**a**) Location of the Oromia Region, (**b**) sketch of the three geographic sectors of the Main Ethiopian Rift, in the blue square, the studied area (**c**) geological map of the study area (modified from [41]).

The extensional tectonics are responsible for the presence of lineaments that are subparallel to the NE–SW-trending rift axis. In particular, at least three sets of faults have been recognized in the area: NNE–SSW, N–S, and NNW–SSE trending faults. The first two sets are dominant and extend for long distances, following the axis of a tectonically active fault system called the Wonji Fault Belt (WFB). In addition, E-W- to NW-SE-trending, cross-rift oblique-slip fault systems have been locally observed [38,39]. The marginal escarpments have well-defined, steep normal-fault scarps [40].

Older structural trends have been identified in the highlands, where faults with different orientations are reported [34,41].

Active faulting within the rift (i.e., Wonji Fault Belt) causes geothermal and fumarolic activity, and high-temperature thermal springs occur along the border of some lakes and in connection with scattered volcanic centers [21].

#### **3. Hydrological Setting**

Two major aquifer classes can be identified in the MER: (i) Extensive aquifers with intergranular permeability (unconsolidated sediment: alluvium, eluvium, colluvium, and lacustrine sediment), and (ii) extensively fractured and weathered volcanic rocks (basalts, rhyolites, trachytes, and ignimbrites) [42]. The latter shows variable transmissivity and different hydraulic conductivities in relation to both the degree of fracturing and weathering, and the presence of interbedded palaeosols and alluvial or pyroclastic deposits within the volcanic series [43].

In fractured volcanic rocks, the flow is predominantly fault-controlled and laterally discontinuous: The groundwater flows parallel and subparallel to the main trending faults, influencing the subsurface hydraulic connection among the rift lakes and the relationship between the river and groundwater. Local palaeochannels drive the groundwater flow (i.e., the palaeochannel along the Bulbula River, which connects the Ziway and Abiyata Lakes) [40,44].

Based on geological characteristics and structural settings, distinct hydrogeological systems are recognized in the highlands and lowlands [45]. Lowlands are characterized both by fractured basaltic and ignimbritic aquifers, which are mainly unconfined or semiconfined, and permeable alluvial and colluvial deposits or lacustrine deposits forming the main shallow aquifers. Conversely, in the highlands, alluvial deposits (in alluvial plains or strips along river courses) and weathered volcanic rocks (with limited interbedded alluvial gravels and sands) are present, forming multilayer confined, semi-confined, and unconfined aquifers.

In general, the main sources of recharge are rainfall and river channel losses [42]. The rate of recharge is strongly influenced by the distribution and amount of rainfall, the permeability of the rocks, the geomorphological setting, and the presence of surface water close to major unconfined and semi-confined aquifers. The main groundwater recharge occurs in the highlands and preferentially moves along the rift within the large regional faults parallel and subparallel to the axis of the rift. In most cases, the escarpments act as discharge areas, which is manifested by the occurrence of high-discharge and faultcontrolled springs along them. The rift floor represents a regional discharge area, and indirect recharge from rivers and lakes occurs in the highly faulted area. Recharge and groundwater flow in the weathered upper zone of the highlands are the driving force for much of the hydrology of the Zway-Shala basin, rather than deep upwelling flow systems with long residence times. The sharp decrease in elevation favors fast drainage of the groundwater in the form of springs. The springs located along the steep boundary between the rift and escarpment have very high seasonal variations in discharge (Figure 2). The same situation exists in highland areas close to deeply incised and large river valleys [40].

**Figure 2.** Map of groundwater recharge. A = widespread good quality groundwater at a relatively shallow depth (dominantly highland volcanic aquifers recharged by high rainfall). B = large groundwater reserve with fair to bad quality often localized in lower elevation areas (rift valley and volcanics in pediment covered with thick sediments and intermountain grabens), C = low to moderate groundwater reserve with fair quality (highland trap series volcanic aquifer with less sediment cover and recharge), D = medium to high groundwater reserve in the volcanics and sediments recharged by rainfall and rivers in places with serious salinity problem, E = low groundwater reserve with moderate quality recharged by seasonal floods and streams. aa': Profile sketching the hydrogeological setting of the area (modified from [40]). The red circle shows the investigated area.

#### **4. Materials and Methods**

A field survey and groundwater sampling were performed in the districts of Shashemene, Arsi Negelle, and Siraro. The field work consisted of a census of the existing water points (wells and springs). The wells were classified into deeply drilled wells (ranging from 20 to 400 m deep) and shallow manually dug wells (up to 20 m deep).

The location of water points is reported in Figure 3. Corresponding to each water point, information about the location (geographic coordinates) and the water point features (depth, diameter, discharge) was collected. Moreover, a picture showing the conditions of the wells and springs was included. The data collection was realized with the support of the administrative staff of the Woredas and the cooperation of social promoters (LVIA).

**Figure 3.** Types and Distribution of the Water Sampling Points in the Study Area.

To verify the groundwater quality, the following physical-chemical parameters (pH, electrical conductivity [EC], temperature, and fluoride concentration) were measured in situ with field instruments. The equipment for field measurements consisted of:


The collected field data were reported in a specific field card for each censused water point (Annex 1).

A total of 52 water points was sampled: 16 groundwater samples from shallow wells, 17 from deep wells, and 19 from springs. Groundwater from shallow wells was sampled using a polyvinyl chloride (PVC) bailer, 91 cm long and with a diameter of 4 cm. Bailer sampling techniques required a gentle lowering of the instrument into the water column of the well to reduce potential problems due to fluid turbulence and a proper transfer of water from the bottom of the bailer to sample containers. Bailers, when properly used, are an acceptable sampling tool [46,47]. Water sampling in deep wells was realized by means of pumps after at least five minutes of pumping, or until temperature and electrical conductivity EC remained stable, according to [48]. Sampling from springs was performed directly in correspondence with the water emergence from the soil. Groundwater was stored in 100-mL polyethylene bottles.

Water analyses were carried out at the Hydrochemical Laboratory of the Department of Earth Sciences at the University of Turin. Major anions (Br, Cl, F, NO2, NO3, SO4) were analyzed using ion chromatography. The concentrations of Ca, Mg, CO3, and HCO3 were measured by titration. Other ions were analyzed by spectrometry. Fluoride concentration was measured both in situ and in the laboratory.

The chemical analyses were followed by the ionic balance calculation. The level of error was calculated by using the following formula:

Error of ion balance = ((Σcations − Σ anions)/(Σcations + Σ anions)) × 100

An error of up to ±5% was considered as tolerable [49].

#### **5. Results**

The census of the water points in the Arsi Negele, Shashemene, and Siraro districts showed the presence of springs, shallow handmade wells, and deeply drilled wells, with a defined geographical distribution. Specifically, in this study we censused 16 shallow wells, primarily located along the eastern escarpment or along the highlands (PM in Annex 2), 30 deeply drilled wells, mainly located in the lowlands, (PT in Annex 2) and 21 springs, mainly located in the easternmost sector of the study area, particularly close to the eastern escarpment of the old caldera rift of Awasa Lake [50] (S or SAN in Annex 2). One of the censused springs (S15/c1) showed thermal physical-chemical features. Water in shallow wells may be contained in altered, highly fractured, or reworked volcanic rocks, residual soils or eluvial/colluvial deposits, while deep wells are mainly drilled in fractured volcanic rocks, intercepting water from the fractures.

The physical features of water (EC, pH, and temperature) showed an appreciable range of variability (Annex 2). The EC ranged from 57 to 1969 μS/cm, with the lowest values measured in the springs (from 57 to 287 μS/cm) and the highest values in the shallow and deep wells (from 109 to 1453 μS/cm). The pH values ranged from 5.5 to 9 in wells and from 5.5 to 6 in springs, with the only exception of the thermal spring, which was strongly alkaline (pH = 8). The distribution of pH and EC showed a decrease from the center of the lowlands (western-central sector of the study area) to the eastern highlands, with similar values closely aligned along an NNE/SSW trend. The temperature ranged from 17.6 to 30 ◦C, with higher values in the deep wells (from 21 to 30 ◦C, 25.4 ◦C on average), and lower values in the shallow wells (from 17.6 to 24 ◦C, 20 ◦C on average) and in the springs (from 18.2 to 25 ◦C). The thermal spring reaches a temperature of 90 ◦C.

The results of the chemical analyses, reported in Annex 2 and summarized in Figures 4 and 5, identified different hydrochemical facies for the different kinds of well. In the springs and in the shallow wells, water samples can be described as both sodium bicarbonate and calcium bicarbonate, even if a higher concentration of HCO3 was registered in the wells. In the deep wells, on the other hand, a prevalence of sodium bicarbonate facies was observed (with the only exception of p27/C1 well), suggesting the presence of a deep groundwater circulation influenced by ion exchange (i.e., Ca-Na ion exchange).

The Schoeller diagram (Figure 5) shows a similar trend for all of the samples, with the exception of the S15c/1 (i.e., thermal spring), which had high Na, Cl, and HCO3 concentrations and very low Mg content.

In all the performed tests, the error of ion balance was inferior to ±5%, and consequently, it was considered tolerable.

**Figure 4.** Piper classification diagram for different water types in the investigated area.

**Figure 5.** Schoeller classification diagram for different water types in the investigated area.

In the deep wells, despite the low concentrations of Ca and Mg, the concentration of F was usually high. Fluoride exceeding the WHO maximum acceptable concentration (1.5 mg/L) was highlighted in 25% of the samples (69% were from deeply drilled wells, 23% from shallow wells, and 8% from springs). More than 21% of the samples ranged between 0.7 and 1.5 mg/L in fluoride concentration (45% were from shallow wells and 55% from deeply drilled wells). A fluoride concentration between 1.5 and 4 mg/L was observed in the area south of Langano Lake (Arsi Negele District), whereas concentrations higher than 4 mg/L were mainly observed in the Awasa volcanic district, in the westernmost study area

(Siraro District), and southwest of Shala Lake, close to the old caldera rift [50]. The lowest values (less than 0.7 mg/L) were observed along the eastern escarpment and highlands, while the maximum concentrations were in the thermal spring of Awasa (S15/c1) and in the deep well PT15x/c1, located northeast of Awasa Lake, with values of 28.6 mg/L and 13.1 mg/L, respectively (Figure 6). However, with the exception of the thermal one, the censused springs generally showed low salinity and low fluoride concentration. The measured concentrations of fluoride describe a distribution similar to the one observed for pH and EC values, with a decrease from the center of the lowlands (western-central sector of the study area) to the eastern highlands, with similar values closely aligned along an NNE/SSW trend.



**Figure 6.** Geographical distribution of the fluoride values recorded in the investigated area.

The concentration of nitrates was usually below the drinking water standards (50 mg/L, [51]). Both hand-dug shallow wells and deep wells showed low nitrate concentrations, inferior to 15 mg/l (with the only exceptions of PM17/c4, having a nitrate level of 36.5 mg/L and PT27/c1, with a nitrate concentration of 45.6 mg/L). In the springs, on the other hand, the nitrate level was generally higher, often showing a concentration greater than 10 mg/L. However, only one of the springs (S16/c1) presented a nitrate concentration higher than the limits, equal to 70.1 mg/L.

#### **6. Discussion**

The geographical distribution of shallow and deep wells and springs in the Oromia region (Figures 3–6) suggests the presence of two different aquifers, respectively exploited with the different kinds of wells. The maps show, indeed, a preferential location of shallow manually-dug wells and cold springs in correspondence with the eastern escarpment and highlands, while the deep wells and the thermal springs are mainly located in the lowland

areas of the rift. The physical and chemical features of water support the hypothesis of two different water circuits, as also observed by Ayenew [40,45] and Rango et al. [21]. Water samples from the shallow wells and from the cold springs showed, indeed, lower temperatures (17 to 25 ◦C) and Ca–HCO3 or Na–HCO3 hydrochemical facies. In contrast, the hot spring and most of the groundwater samples in the rift displayed a Na–HCO3 fingerprint, with Na and HCO3 concentrations representing more than 80% of the ionic species in the solution. Consequently, it is possible to suggest a compositional change from Ca–Mg/HCO3 to Na–HCO3 along the path of groundwater flow from the highlands to the rift floor.

In accordance with the hydrogeological model proposed by [40], the shallow aquifer is hosted in altered rocks, residual soils, and fluvio-lacustrine deposits and is usually present in the uppermost tens of meters, where water flows in porous media. This aquifer is mainly located in the highland end-rift escarpments and appears more productive than those in the deep escarpments, with a decrease in permeability with the degree of alteration. In contrast, the deep aquifer of the lowlands is hosted in fractured volcanic bedrock where water is intercepted at different depths, generally hundreds of metres below the surface (up to 250–300 m). In the northern part of lowland areas, near to the lakes, however, additional shallow porous aquifers may be hosted in the fluvio-lacustrine deposits, as confirmed by the shallow wells PM7, PM11, PM13, PM15.

Data obtained from chemical analyses showed that the increase in Na observed in the deep aquifer affected the F concentration (Figure 7). Similarly, water with relatively low contents of Ca and Mg had a higher concentration of fluoride. High fluoride concentration values reflect a strong removal of Ca from the solution and a concurrent enrichment in Na. An increase in EC and temperature were also associated with the F concentration: The water temperature generally increased with the depth of the wells, as well as with the fluoride concentration (Figure 8).

**Figure 7.** Scatter plot between sodium and fluoride (linear trend with equation *y* = 0.04*x* − 0.7 and R2 = 0.8).


**Figure 8.** Geographical distribution of ce values recorded in the investigated area.

The decreasing trend of fluoride and pH and EC values from the center of the lowlands to the eastern highlands is consistent with the geological and structural framework of the area, in accordance with the type of bedrock, the distribution of surface deposits, and the orientation of the main faults (NNE/SSW to NE/SW trend).

For the springs, various hydrochemical features were recorded, suggesting two different origins, which are also reported in the literature. Springs can have a superficial flow circuit with low electrical conductivity and fluoride concentration, and relatively high nitrate levels, or a deep flow circuit of thermal water with high electrical conductivity, fluoride concentration, and temperature, and low nitrate levels. These data support an anthropic origin of nitrate concentration, probably due to the oxidation of nitrogenous waste products in human and animal excreta [52–54]. The discrimination between animal manure and domestic sewage was not possible using the current data. However, a groundwater isotopic analysis (i.e., using nitrate and boron isotopes) could be a valuable help to distinguish between these two nitrate sources and to implement management actions for groundwater protection [55,56].

The fluoride content has a geogenic origin, which is supported by the geological setting and the low anthropogenic pressure in the region, as well as by the low contents of nitrate and sulfate. During magmatic processes, fluoride does not easily fit into the crystal structure of minerals and is preferentially partitioned into melt, remaining in the residual magma until its crystallization. Consequently, fluoride is more concentrated in evolved sialic rocks, such as the bedrock of the lowland Oromia Region, rather than in primitive basic rocks (e.g., basalts) [22]. The increase in pH and temperature favors the interaction between sialic rock and water because they influence the reactions of dissolution and adsorption-desorption, which contribute to the mobility of the fluoride contained in the primary mineral phases (such as fluoride, apatite, amphibole, biotite, and volcanic glass) or secondary minerals (such as Fe−, Al− and Mn− oxides and hydroxides, and clay minerals), particularly in arid or semi-arid climates [21,57,58].

Locally, the leaching of fluoride in groundwater is connected to the presence of active faults where the circulation of CO2 enhances the dissolution process of silicates, oxides, and hydroxides [59]. In the study area, which is located in an active rift zone, similar conditions are common and can contribute to fluoride mobilization from rock and sediment into groundwater. The alignment of water points (springs and wells) with similar F− ion concentrations along preferential NNE-SSW trends (consistent with the orientation of the main faults) highlights the possible tectonic contribution to the high content of fluorides in the groundwater.

#### **7. Conclusions**

In arid and semi-arid areas, human health and economic development depend on water availability, which can be strongly compromised by droughts [60]. Many people have no access to fresh water. Surface water and precipitation are insufficient, and in some cases, the quality of the water is compromised by natural contaminants.

The present research analyzed the physical and chemical characteristics of groundwater in a rural area of the central MER, particularly in the districts (Woredas) of Shashemene, Arsi Negelle, and Siraro, to create awareness of the availability, quality, and risks connected to water resource management. The study area is located in a symmetrical rift basin where groundwater flows from the highlands to the rift valley, across well-defined marginal faults that form escarpments, and into the fractured rift floor, which is dotted by volcanic cones. The study was developed using a census of the main water points (springs and wells) in the area with sampling and physical-chemical analysis of the water samples, with particular regard to the fluoride concentration. The hydrochemical analyses of the groundwater showed large variability in both geographical location and depth of the water points. Regarding the geographical distribution, most springs and shallow wells were located in relation to the eastern highland and at the bottom of the eastern plateau slopes. The deep-drilled wells were, instead, concentrated in the center lowlands of the rift valley. The depth of the groundwater surface varied within a wide range. The depth of the groundwater surface ranged from 250 to 300 m in the lowlands to a few meters in local shallow aquifers with scarce productivity, whereas, in most of the highland plains, the water depth did not exceed 50–60 m. Fluoride represents the main pollutant of groundwater. The highest fluoride contents were measured in the deep-drilled wells of the lowlands, which were related to the fractured aquifer, or in the thermal springs. Moreover, similar fluoride concentrations were primarily aligned in an NNE-SSW direction, similar to the orientation of the main fault systems of the MER. Low fluoride contents (<0.7 mg/L) were normally measured in connection with the shallow and porous aquifer of the highlands and eastern escarpment. Commonly, high fluoride concentrations were associated with Na–HCO3-type water. Two different origins were observed for the springs: A superficial flow circuit with low electrical conductivity and fluoride concentration and relatively high nitrate levels, or a deep flow circuit of thermal water with high electrical conductivity, fluoride concentration and temperature, and low nitrate levels.

The study describes the hydrogeological setting of the study area, providing a framework of the water-point distribution and water quality. The collected data provide information on the risks with rural communities and local governments. Moreover, these results can be useful to support the adequate use of the available water resources and to plan proper actions to increase access to fresh water (new water points, locations, expected depths of the water level, associated costs, etc.), aimed at the sustainable human and rural local development of the region.

Future studies in this region may include the measures of the static level of the water table in different periods of the year in order to evaluate the influence of the seasonality (i.e., dry season and rain season) on the water resource. In addition, quantitative analysis through water flow measurements would be important to assess the real amount of available water in correspondence with the different types of wells.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2071-105 0/13/3/1347/s1.

**Author Contributions:** Conceptualization, S.M.R.B. and M.L.; methodology, S.M.R.B. and M.L.; software, M.L., C.C.; validation, S.M.R.B., M.L., D.A.d.L.; formal analysis, C.C.; data curation, C.C.; writing—original draft preparation, C.C., S.M.R.B., M.L.; writing—review and editing, C.C.; supervision, D.A.d.L.; project administration, S.M.R.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data is contained within the article or supplementary material.

**Acknowledgments:** The authors desire to thank LVIA Onlus for the useful contribution in the on-site activities.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Article* **Development of a Real-Time, Mobile Nitrate Monitoring Station for High-Frequency Data Collection**

**Martin Jason Luna Juncal <sup>1</sup> , Timothy Skinner 1, Edoardo Bertone 1,2,3,\* and Rodney A. Stewart 1,2**


Received: 29 May 2020; Accepted: 16 July 2020; Published: 17 July 2020

**Abstract:** A mobile monitoring station was developed to measure nitrate and physicochemical water quality parameters remotely, in real-time, and at very high frequencies (thirty minutes). Several calibration experiments were performed to validate the outputs of a real-time nutrient sensor, which can be affected by optical interferences such as turbidity, pH, temperature and salinity. Whilst most of these proved to play a minor role, a data-driven compensation model was developed to account for turbidity interferences. The reliability of real-time optical sensors has been questioned previously; however, this study has shown that following compensation, the readings can be more accurate than traditional laboratory-based equipment. In addition, significant benefits are offered by monitoring waterways at high frequencies, due to rapid changes in analyte concentrations over short time periods. This, combined with the versatility of the mobile station, provides opportunities for several beneficial monitoring applications, such as of fertiliser runoff in agricultural areas in rural regions, aquaculture runoff, and waterways in environmentally sensitive areas such as the Great Barrier Reef.

**Keywords:** agriculture; Nitrate runoff; real-time monitoring; water quality

#### **1. Introduction**

Nitrate fluxes in the water system due to fertiliser mismanagement create substantial environmental and health issues where, particularly in rural communities, efficient and reliable water treatment processes are not always available [1]. Reliable monitoring of nitrates is therefore critical to understanding and predicting nutrient runoff and, in turn, poor fertiliser application practices.

Traditional methods rely on time-intensive sampling and analysis processes, usually through surface water samples which are collected and analysed in a laboratory (Table A1, Appendix A) [2,3]. Real-time nutrient sensors have the potential to generate widespread benefits for agriculture and aquaculture industries, while also providing a rapid source of high-frequency data that can assist in mitigating environmental damage in sensitive areas.

One such method of automation involves the use of high-frequency optical sensors. However, the reliability and potential successes of optical nitrate sensors is understudied and has been questioned previously [4]. One project relied on real-time sensors for pH, conductance, temperature, turbidity, chlorophyll and dissolved oxygen monitoring in water streams; however, to complete nutrient analyses, manual sampling was required to develop a regression model that could estimate bacterial concentrations [5]. As a result, the study may have benefited from the use of real-time nutrient sensors

due to the potential ability to correlate temporal data with the water quality probe results. Another study implemented real-time monitoring buoys to obtain nutrient data [6]. While innovative, these data are limited due to the small measurement range of the instrument (0–5 mg L−<sup>1</sup> of NO3), as well as a maximum autonomous monitoring period of six months [7].

Currently, there is not a complete acceptance of real-time nutrient sensors for reliable mainstream water quality monitoring [8]. As a result, this perceived unreliability leads to skepticism, which in turn hinders opportunities for the widespread implementation of real-time nutrient sensors. However, previous studies [9] have shown that readings from optical sensors can be compensated for optical interferences while simultaneously deployed remotely. As a consequence, there is the potential, through a combination of targeted experiments and data-driven modelling, to improve the reliability of in-situ optical sensors, such as those targeting nitrates. In turn, this would assist in allowing responses to environmental issues to occur in real-time.

Consequently, the overarching goal of this study was to design, construct, calibrate and field test a real-time nitrate measurement instrument to enable high-frequency and remote nitrate monitoring applications to occur in the future. This goal was achieved through a number of research tasks, including comparing the accuracy of a NiCaVis 705 IQ sensor (from YSI, Yellow Springs, OH, USA) to a spectrophotometer, under laboratory conditions. The laboratory experiments also contributed to the second aim, which was to design a compensation model for nitrate optical sensors. Finally, this testing and the subsequent compensation models contributed to the third aim of the study, which was to develop and deploy a mobile monitoring station to obtain and transmit water quality data, especially nitrate data, in real-time and at very high-frequencies (30-minute intervals). The achievement of these three aims contributed to our goal to foster the deployment of reliable real-time nitrate sensors for a range of important applications.

#### **2. Materials and Methods**

#### *2.1. Phase I. Laboratory Calibration*

To start the development of a mobile monitoring system, a comprehensive laboratory calibration of the NiCaVis 705 IQ sensor was conducted. This sensor specialises in identifying the concentration of several water quality parameters in real-time, at high-frequency intervals. The focus of this particular study was on calibrating its nitrate detection capability [10]. As a result, the sensor was specifically calibrated to account for the expected waterway conditions of the region, which are likely to differ significantly depending on the location of deployment. However, most rivers tend to share similar water quality ranges; hence, the goal was to develop a compensation range that could make the nitrate readings from NiCaVis reliable and consistent with what is typically observed in these waterbodies [11,12].

To achieve this, potassium nitrate solutions with concentrations ranging between 0.05 mg L−<sup>1</sup> and 500 mg L−<sup>1</sup> were created, thus allowing a set of samples capable of simulating eutrophication conditions (concentrations > 2 mg L−<sup>1</sup> [13]) to be developed, while also enabling the assessment of the efficacy of the NiCaVis 705 IQ sensor through a broad enough concentration spectrum [14–18].

The preliminary calibration and analysis was conducted by measuring potassium nitrate solutions with the NiCaVis 705 IQ sensor as well as a laboratory spectrophotometer and comparing the results as a standardisation measurement. Upon developing this baseline for both devices, interference sources consisting of alterations to the pH, salinity, turbidity, humic acid content, bromide content and temperature were made to calibrate the sensor under different conditions, particularly since studies have shown that these parameters may reduce the accuracy of optical sensors [9,19–22]. Thereafter, a repetitive measurement method was followed to obtain the sought-after results. Firstly, 50 mL of potassium nitrate was placed in tubes, where an interference source was applied and thoroughly mixed to create a homogenous solution. Each sample was then analysed by a laboratory spectrophotometer and the NiCaVis 705 IQ, and the results were compared afterwards. Finally, the remaining liquid was discarded, and the process was repeated for the remaining interferences.

Upon completing the laboratory calibration, the data were analysed to validate the sensor's readings and to develop compensation models as required. In doing so, a multiple regression model was developed to predict real nitrate concentrations based on two input variables (turbidity and raw nitrate readings).

#### *2.2. Phase II. Mobile Trailer Development*

The development of the mobile trailer was conceptualised to address some of the issues regarding the accessibility and implementation of traditional water quality monitoring methods. Sampling from rural locations at regular intervals is costly, as is the implementation of methods to maintain the health of the waterways in these isolated areas. Thus, a mobile trailer was considered useful due to its movability to different locations, as well as its capability to remotely transmit nitrate and other water quality data via a cloud-based system, thus eliminating the need for continual trips to rural areas.

However, to develop the trailer, two aspects had to be considered: the exterior casing and the interior space. For the exterior, the aim was to obtain a relatively inexpensive trailer that had the capability to be adapted and modified to suit the project's needs, while the interior required sufficient spacing to hold two sensors, an analysis container, communications/pumping systems, as well as additional auxiliary electronics. Figure 1 displays the exterior modifications, where solar panels were added to a trailer and connected to batteries to provide sufficient power supply to the system. Figure 1 also depicts the overall housing for the sensors, where a steel-cased trailer was selected to ensure that a robust, durable system could be setup.

**Figure 1.** Trailer exterior modified with solar panels.

Figure 2 then displays the interior design that was adopted to house the electronics, communication system and sensors.

**Figure 2.** Interior spacing of the trailer, showing the wet area, communications box and sensor equipment.

Overall, Figures 1 and 2 display relatively simple housing for the required systems, where an additional compartment was used to store a pumping system. The design of the mobile trailer was created with the intention that the entirety of the water quality monitoring instruments could be held within the one compartment. Therefore, only a small pump had to be placed outside of the trailer to extract water into the wet area, where the sensors could identify the nitrate and water quality parameter concentrations at half-hourly intervals. The data were then streamed through the communications datalogger and automatically uploaded online to a cloud-based server, known as Eagle.IO (https://eagle.io/). Further details regarding the piping and instrumentation layout are displayed in Figure A1 of Appendix B.

Overall, the mobile trailer was designed and constructed to be robust and reliable for field research applications. While the design could be refined further and optimised, the focus of the current study was on its capability and accuracy to remotely collect high frequency water quality parameter data. Future design optimisation work could be conducted on the trailer shelter, water sample pumping system or the renewable energy system, where reduced sizes and capacities could result in cost-savings without compromising operational performance.

#### *2.3. Phase III. Pilot Field Study*

The pilot field study took place at Hilliards Creek in Queensland, Australia. Located in the small town of Ormiston (population < 6000), the site provided an ideal opportunity to test the monitoring trailer in a secure and secluded area. Figure A2 in Appendix C displays a map of the pilot location, which was chosen due to the site's proximity (3.4 km) to an upstream wastewater treatment plant (WWTP). The WWTP is known to discharge nitrate-containing effluent into Hilliards Creek, and hence the location provided the opportunity to achieve this study's goals by providing a location to collect accurate, high-frequency nitrate data.

In order to collect the water quality data, the sampling pump was placed in the water at approximately 30 cm above the sediment bed. Agrometeorological data were collected at half-hourly intervals, which included nitrate concentrations (from the NiCaVis 705 IQ), as well as the pH, turbidity, salinity and temperature (obtained by Xylem Analytics' EXO2 Sonde, from YSI, Yellow Springs, OH, USA) of the waterway. Furthermore, rainfall data were obtained at the same frequency from a gauge located adjacent to the WWTP. Prior to the fully automated implementation, the system was calibrated using surface water samples, which were verified with laboratory water quality analyses. The trailer was then left to collect data for nearly four days, before a 1% average exceedance probability (AEP) rain event halted further monitoring due to disruptions from a subsequent flash flood. Despite this, the data were collected through Eagle.IO, where no on-site monitoring was needed after the initial calibration of the equipment.

#### **3. Results**

#### *3.1. Phase I. Laboratory Calibration*

Figure 3 illustrates a comparison between the nitrate readings obtained with the NiCaVis 705 IQ sensor and a laboratory spectrophotometer, for concentrations ranging between 0.05 mg L−<sup>1</sup> and 500 mg L−<sup>1</sup> without interference sources applied.

**Figure 3.** Expected concentration from known solutions vs. measured nitrate concentrations, NiCaVis 705 IQ sensor and laboratory spectrophotometer. Initial calibration data.

For the NiCaVis 705 IQ sensor, good accuracy was achieved up to a concentration of 25 mg L<sup>−</sup>1, with a rapid drop in accuracy thereafter. The sensor was unable to measure nitrate concentrations beyond 250 mg L<sup>−</sup>1. The spectrophotometer was shown to be accurate for nitrate concentrations only up to 3 mg L<sup>−</sup>1; readings beyond this showed a constant concentration of 5 mg L−1. This was believed to have resulted from an Inner Filter Effect (IFE), which occurred as a result of the analyte concentration being too high, causing signal loss in the spectrophotometric analysis [23]. Additional figures to emphasise the performance of the sensor and spectrophotometer when measuring interference sources such as turbidity, pH, temperature, bromide, humic acid and salinity are presented in Appendix D. The results of these experiments display a similar trend to the one observed in Figure 3, emphasising the higher accuracy of the sensor when compared with traditional water quality monitoring methods.

Overall, NiCaVis was able to determine concentrations within 0.1 mg L−<sup>1</sup> of the baseline in 81% of the samples tested (*n* = 240), while the spectrophotometer was only able to achieve this in 63% of tests. Furthermore, the sensor read concentrations within a 0.5 mg L−<sup>1</sup> range in 96% of cases, while the spectrophotometer had the same result in only 84% of samples. If accounting for the error in samples >5 mg L<sup>−</sup>1, the number of samples within that range of accuracy decreases to 67%.

This study also found that the accuracy of NiCaVis can be improved through extensive compensation modelling and reading adjustments. Turbidity was found to be the largest interference source: elevated levels of turbidity with lower nitrate concentrations had the tendency to reduce the sensor's ability to accurately detect the level of nitrate. Figure 4 displays this correlation, indicating that a smaller nitrate presence causes a larger signal loss with increasing turbidity. Figure 5 then displays the relationship between the expected and experimental nitrate concentrations, based on different turbidity levels.

**Figure 5.** Expected vs. experimental nitrate concentrations for varying turbidity.

From the above data, a preliminary compensation model was developed, as shown in Equation (1) (*R*<sup>2</sup> > 0.99).

$$\text{[NO}\_3\text{A]} = 0.961 \, ^\circ \text{[NO}\_3\text{S]} + 0.003 \, ^\circ \text{Tb} - 0.119\,,\tag{1}$$

where [NO3,A] is the actual corrected nitrate concentration of the sensor, [NO3,S] is the nitrate concentration read by the sensor and Tb is the turbidity of the waterbody.

This research found that there were limitations to the model, and hence further analyses would be necessary to reinforce the validity of interference compensations, while additional studies could help to develop more complex threshold models to better determine the corrected nitrate reading based on more refined intervals. Regardless, the model was practical in correcting erroneous results for nitrate concentrations >0.1 mg L−1; however, significant inaccuracies were noted for lower concentrations. Despite this, the likelihood of having concentrations <0.1 mg L−<sup>1</sup> in farmland rivers or creeks susceptible to nutrient releases is low, which is evidenced by the real-time data collected during the pilot study (Figure 6).

**Figure 6.** Nitrate–rainfall data at Hilliards Creek, 14th to 15th January 2020.

#### *3.2. Phase II and III. Mobile Trailer Development and Pilot Field Study*

Figure 6 shows the initial relationship between nitrate concentrations and rainfall over the pilot study monitoring period, based on the data extracted from Eagle.IO. Figure A20 in Appendix E displays the virtual graphing interface that the software uses to visualise readings collected by the sensors.

Figure 6 highlights the nitrate–rainfall correlation for the operational cycle of the NiCaVis sensor until it was turned off for maintenance. Furthermore, approximately 12 hours of rainfall preceding the NiCaVis' commencement of recording data is also shown.

The data seem to show that rainfall events caused delayed decreases in nitrate concentrations due to the dilution of water entering the creek. Such a delayed dilution effect was expected due to the inherent processes involved with runoff generation following rainfall, as well as flow movement downstream. In addition, given that the location of the rainfall gauge is 3.4 km upstream from the trailer deployment site, this further emphasised why the delay between rainfall and nitrate readings existed. Although the lack of nitrate data for the 14th of January does not allow us to validate such claims, the dilution effect portrayed in Figure 6 is also supported historically, with Figure 7 showing the variation of nitrate concentrations that have occurred seasonally from 2003 to 2014.

**Figure 7.** Historical Hilliards Creek nitrate data from May 2003 to May 2014.

Figure 7 displays an evident trend where nitrate concentrations had the tendency to lower significantly during the wet season (November to April) and rise during drier periods. Therefore, this

data supports the results observed in Figure 6, displaying the inverse correlation between rainfall volume and nitrate concentration in waterways.

In addition, Figure 7 also presents a significant difference between the nitrate concentrations at Hilliards Creek (recorded 200 metres downstream from the current study) and at a location further downstream (3.61 km away). As a result, comparing the historical monthly upstream data to the results highlighted in Figure 6 indicates that an inflow source (the WWTP) was causing significant increases in nitrate concentrations, which then diluted downstream.

Furthermore, the variation of the data in Figure 6 also emphasises the necessity of real-time and high-frequency nutrient monitoring as significant variations are possible in short timespans. The data in Figure 7 show significant historical variations on a monthly basis and hence indicates that the typical sampling frequency (i.e., monthly) used for monitoring is unlikely to be truly representative of a waterway's health or of shorter-term nutrient fluctuations, which appear to occur.

Figure 8 displays the turbidity data collected by the Xylem Analytics' EXO2 Sonde, coupled with rainfall data.

**Figure 8.** Turbidity–rainfall data at Hilliards Creek, 14th to 18th January 2020.

The correlation between preceding rainfall and turbidity is evident due to a lag of eight hours between the commencement of the rainfall event and the first increase in turbidity. In addition to causing an optical interference on nitrate readings, it is possible that more turbid waters may reduce the ability of chlorophyll-*a* to absorb energy from the sun, in turn reducing algal blooms and preventing nutrient uptake [24]; hence, simultaneous monitoring of turbidity and nitrates is critical for accuracy. Further analyses would be required to affirm this correlation; however, collecting more consistent nitrate, chlorophyll-*a*, turbidity and rainfall data could reveal some patterns that may be modelled and integrated into future studies to predict and manage the physicochemical conditions of waterways.

At 5:30PM on 16/01/2020, the mobile trailer system was turned off to complete operational maintenance. When the system was re-established, a significant rainfall event occurred afterwards, thus allowing further correlations to be made, as illustrated by the data presented after 12:30PM on 17/01/2020. Consequently, Figure 8 displays significant turbidity rise associated with the overnight rainfall event that occurred between 3:00AM and 5:00AM on January 18, 2020. This sudden and large burst of precipitation, noted as a one-in-100-year event, caused an increase in turbidity from 1.3 NTU (which was the base level detected with no preceding rainfall), to 19.33 NTU, an almost 15-fold increase. However, due to the severity of this event, elevated creek levels occurred such that the mobile system was inoperable for several days thereafter.

Despite this, the data presented in Figure 8 showed that rainfall events, particularly those of large magnitudes, may cause significant elevations in turbidity, and hence the data would require compensating to ensure that the readings made by the NiCaVis 705 IQ sensor and its auxiliary units are maintained at an accurate level.

#### **4. Discussion**

The completion of the pilot study consisting of laboratory calibrations and compensation modelling, as well as the trailer development and its subsequent deployment, has the potential to provide numerous benefits. The advantages of obtaining high-frequency nutrient data were explored in this study, where the 30-minute interval results provided the opportunity to observe the highly dynamic nature of water quality parameters. Consequently, the next sections briefly discuss the potential applications of the mobile trailer, as well as the implications of collecting data at a high frequency over long time-periods.

#### *4.1. Agricultural Benefits in Rural Regions*

The overproduction of crops to meet increased food demand has caused increases in fertiliser usage, leading to significant environmental damage and exemplifying the need for real-time, remote monitoring at vulnerable waterways. For lower socioeconomic nations, such as those within the tropics, routine monitoring through traditional methods is not feasible due to their diminished capability to proportionately improve their agriculture production [25]. As a result, countries in the tropics are struggling to implement monitoring regimes that can mitigate environmental harm and minimise economic losses from mismanaged fertiliser usage. Thus, one such application of a real-time, high-frequency system for water quality monitoring is to provide vulnerable nations with a way to continually monitor agricultural effluent. In doing so, fertiliser runoff can be effectively managed using real-time monitoring, thus providing an option to assist decision-makers in the reduction of economic losses associated with high-volume discharges.

#### *4.2. Minimisation of Nutrient Pollution from Aquaculture Practices*

Aquaculture industries pose a significant threat to waterway health. In recent years, nutrient inputs have significantly declined, with the conversion efficiency to assist in the growth of aquatic organisms also improving [26]. However, excretion products from these organisms persist and have the potential to cause significant environmental and ecosystem changes. This therefore presents issues where excess nutrient loads can not only lead to eutrophication but can cause disruptions to marine species. One study [27] found that up to 45% of nitrogen provided as a food source could be excreted by some organisms, emphasising the significant detriment that mass aquaculture production may have on water quality. Therefore, real-time, high-frequency monitoring can be applied to better equip stakeholders to manage effluent releases into general waterways. In doing so, the timing and volume of nutrient-rich discharge can be managed to provide a lower overall threat to waterways and ecosystems.

#### *4.3. Improvement of Environmental Health in Environmentally Sensitive Areas*

As previously mentioned, traditional water quality data are usually obtained, analysed and reported at monthly intervals. This minimises labour and technological costs while also providing relatively accurate results regarding the health of waterways. However, this study has highlighted that the overall complexity of river/creek systems is high, with the dynamic nature of these systems causing large fluctuations of physicochemical and nutrient concentrations in short time-periods. For typical waterbodies, the importance of measuring the significance of these variations may not be particularly necessary; however, some regions are very susceptible to small changes in water quality. The Great Barrier Reef (GBR) is an example of this, and it is an environmentally sensitive area; coral bleaching is a well-known issue that is associated with changes in water temperature, nutrient levels (particularly nitrate) and lighting [28,29]. As a result, this presents an opportunity for high-frequency monitoring, which has already been conducted and validated as part of this study. Monthly observations of areas such as the GBR are insufficient for maintaining the health of these regions, and hence the potential to

obtain real-time data that can be readily acted on has the potential to mitigate environmental damage to vulnerable areas.

#### *4.4. E*ffi*cacy of Real-Time, High-Frequency Sensors for Routine Water Quality Monitoring*

Traditional high-frequency monitoring systems are too immobile to assist in targeted water quality analyses [8]; however, the mobile system developed in this study has the capability to be deployed at different targeted locations on-demand.

The accuracy of high-frequency optical sensors has been previously questioned [8]. Efforts are being made to improve the reliability of implementing such devices for prolonged use; however, compensating readings accurately has also proven to be beneficial, as a study by de Oliveira et al. [9] has suggested. This study further emphasised this significance by obtaining data to develop a turbidity-nitrate compensation. The calibration results shown in Figures 4 and 5 are in line with the findings of de Oliveira et al. [9] since an increase in turbidity resulted in a decrease in signal strength of the real-time optical sensor. However, we also proved that such a decrease is not only proportional to turbidity concentration but also to the actual nitrate concentration, with low nitrate levels being more susceptible to turbidity interferences and thus a loss in sensor accuracy. Our compensation model extended from [9] to include this second key variable.

#### **5. Conclusions**

Due to a lack of resources and effective water monitoring methods, nations in the tropics are at a greater risk of waterway pollution from excess nutrient loads. Current monitoring methods for nitrates are outdated and fail to deliver consistent solutions to water quality issues in river/creek systems. Furthermore, accessibility and affordability limitations of monitoring systems are significant drawbacks to routine checking of agricultural waterways in rural areas. As a result, a mobile monitoring station was developed in this study to collect high-frequency nitrate and physicochemical water data. The implications of this effective and practical monitoring system, which can be reliably compensated, are significant. Throughout this study, the overall user-input requirements of the monitoring system have been greatly optimised; wireless internet streaming and a nearly complete remote functionality have provided the opportunity for efficient and constant data transmission to occur with minimal human involvement. Furthermore, by developing the streaming platform through code-based software, the necessity for human interactions with the monitoring system, aside from the initial setup and calibration, is minimal.

This study has also identified several findings relating to the modernisation of water quality monitoring. Firstly, extensive laboratory calibrations linked interference sources including pH, turbidity, temperature, bromide, salinity and organic matter, to the potentially reduced performance of a real-time optical nitrate sensor. In doing so, it was found that turbidity had the greatest effect, and hence a compensation model was developed to maintain the accuracy of reported results, even in turbid waterway conditions. The benefits of this process are potentially significant: the reliability of real-time, high-frequency sensors considerably improves through such modelling, where the simultaneous employment of a remote sensor, linked to an online cloud-based data transmitter, can provide accurate data based on compensations associated with environmental conditions. Furthermore, these calibrations have also shown that real-time sensors can be significantly more accurate than traditional laboratory spectrophotometers, thus highlighting the necessity of modernising water quality monitoring methods to minimise time wasted and improve the accuracy and reliability of reported data. Finally, this study has shown how crucial high-frequency measurements are when monitoring water quality as several parameters, particularly nitrates, can have very short-term fluctuations in concentration. Given waterway susceptibility to eutrophication, such erratic variances can have significant consequences, and hence this study was important to further highlight the necessity of high-frequency monitoring.

Future work will expand on the research described in this paper, and remote, live monitoring and advanced data analytics will be coupled to create a fertiliser management decision-support system that can be deployed to combat targeted issues such as fertiliser dosage mismanagement and aquaculture-based nitrate monitoring. Not only would this provide farmers with an on-demand method of managing the water quality of rivers surrounding their sites, but they can also optimise fertiliser dosages and nutrient feeds for a variety of crop types and aquatic organisms to mitigate waterway pollution. Optimal fertiliser usage pays off economically for farmers due to reduced annual demand, while aquaculture monitoring assists in preventing ecosystem damage to critical species and environments.

To further refine the concept, additional trailers at different points along a targeted waterway would provide additional opportunities to verify the concentrations of nitrate from farmland areas. This study has already begun to highlight different factors that affect the severity of nitrate runoff; however, more continual high-frequency data would provide significant opportunities to interrelate these parameters. Overall, more in depth studies are required to develop specific monitoring regimes with high-frequency data. Despite this, the current pilot monitoring program has shown potential, and hence the success of future projects could provide significant benefits to agricultural stakeholders and aquaculture locations, as well as environmentally sensitive areas, particularly in more vulnerable regions, which are susceptible to the effects of overpopulation and climate change.

**Author Contributions:** Conceptualization, M.J.L.J., E.B., R.A.S.; methodology, M.J.L.J., E.B.; software, M.J.L.J., E.B; validation, M.J.L.J., E.B.; formal analysis, M.J.L.J., E.B.; investigation, M.J.L.J., T.S.; resources, T.S.; data curation, M.J.L.J.; writing—original draft preparation, M.J.L.J.; writing—review and editing, E.B., R.A.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** We are thankful to Xylem Analytics Inc. and TEW for their in-kind contribution of several key instruments, materials and resources which were used in this project. Furthermore, we would like to acknowledge the Griffith Scientific Services Laboratory for allowing us to use their facilities throughout the calibration phase, in-kind. We would also like to thank Lawrence Hughes (Griffith University) for his assistance with the monitoring station's design. Finally, we would like to thank Ian Underhill and his staff at the Griffith University Engineering Laboratory for providing us with advice and facilities to assist in developing our monitoring station.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A. Summary of Typical Water Quality Measurement Methods**


**Table A1.** Advantages and disadvantages of different nitrate measurement methods.


**Table A1.** *Cont.*

Table A1 shows that several methods have significant disadvantages, which may minimise the reliability, accuracy or affordability of regular monitoring due to their dependency on manual sampling, in-situ requirements or the time required to obtain accurate data. Laboratory spectrophotometry is commonly used in industry water quality monitoring programs, particularly by government organisations who use the data to assist in decision-making regarding targeted waterways. However, this method also has limitations where it largely relies on water quality scientists to extract samples for analysis, which may not be feasible in rural areas due to the significant distance from city centres and monitoring laboratories [32,33].

#### **Appendix B. Piping and Instrumentation Diagram of the Mobile Trailer**

A piping and instrumentation diagram is shown in Figure A1 to reiterate the functionality of the monitoring trailer, emphasising how the design was developed to obtain water quality data.

**Figure A1.** Piping and instrumentation diagram of the monitoring trailer's functionality.

### **Appendix C. Map of the Study Site in Ormiston, QLD**

**Figure A2.** Preliminary field site location at Hilliards Creek, Ormiston [34].

#### **Appendix D. Additional Figures Comparing the NiCaVis 705 IQ to a Spectrophotometer under Di**ff**erent Interference Conditions**

Generally, the data shown in this section highlights the efficacy of the NiCaVis 705 IQ sensor and indicates that it is much more consistent in the results it provides under a range of conditions when compared with the spectrophotometer results. Overall, the spectrophotometer was unable to accurately identify nitrates whose concentrations exceed 5 mg/L due to a saturated absorbance and the production of nitrites. For the NiCaVis 705 IQ sensor, the accuracy dropped after 25 mg/L due to the Inner Filter Effect (IFE) causing the signal to decrease.

**Figure A3.** Comparison between the expected results and the spectral/sensor data under standard conditions.

**Figure A4.** Comparison between the expected results and the spectral/sensor data with 10 mg L−<sup>1</sup> of bromide added.

**Figure A5.** Comparison between the expected results and the spectral/sensor data with 20 mg L−<sup>1</sup> of bromide added.

**Figure A6.** Comparison between the expected results and the spectral/sensor data with 5 μL per 50 mL potassium chloride (58,670 μs cm<sup>−</sup>1) salinity added to the samples.

**Figure A7.** Comparison between the expected results and the spectral/sensor data with 10 μL per 50 mL potassium chloride (58,670 μs cm<sup>−</sup>1) salinity added to the samples.

**Figure A8.** Comparison between the expected results and the spectral/sensor data with samples at pH 4.5.

**Figure A9.** Comparison between the expected results and the spectral/sensor data with samples at pH 10.

**Figure A10.** Comparison between the expected results and the spectral/sensor data with samples at 4.3 ◦C.

**Figure A11.** Comparison between the expected results and the spectral/sensor data with samples between 35–45 ◦C.

**Figure A12.** Comparison between the expected results and the spectral/sensor data with 1 NTU turbidity added.

**Figure A13.** Comparison between the expected results and the spectral/sensor data with 2 NTU turbidity added.

**Figure A14.** Comparison between the expected results and the spectral/sensor data with 5 NTU turbidity added.

**Figure A15.** Comparison between the expected results and the spectral/sensor data with 10 NTU turbidity added.

**Figure A16.** Comparison between the expected results and the spectral/sensor data with 50 NTU turbidity added.

**Figure A17.** Comparison between the expected results and the spectral/sensor data with 100 NTU turbidity added.

**Figure A18.** Comparison between the expected results and the spectral/sensor data with 10 mg L−<sup>1</sup> of humic acid added.

**Figure A19.** Comparison between the expected results and the spectral/sensor data with 20 mg L−<sup>1</sup> of humic acid added.

#### **Appendix E. Eagle.IO Online Interface**

Figure A20 illustrates the online Eagle.IO interface, which was used to view data trends and identify any potential errors with the system's operation.

**Figure A20.** Eagle.IO online interface.

The interface provides the data on a fixed scale, with Figure A20 highlighting initial nitrate concentration trends on a −1 to 1 scale. From a data analysis perspective, the Eagle.IO interface was not useful in assisting to identify how consistent the reporting and results were. The data were downloaded in a.csv format for further and more detailed analyses.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **A Vulnerability Assessment in Scant Data Context: The Case of North Horr Sub-County**

#### **Velia Bigi 1,\* , Alessandro Pezzoli 1, Elena Comino <sup>2</sup> and Maurizio Rosso <sup>2</sup>**


Received: 28 May 2020; Accepted: 14 July 2020; Published: 27 July 2020

**Abstract:** In Kenyan rural areas belonging to the Arid and Semi-Arid Lands (ASALs), water quantity and water quality are major issues for the local population. In North Horr Sub-County water quality is threatened by nitrate contamination due to fecal matter pollution. This research, hence, aims at assessing the vulnerability of open shallow water sources to nitrate contamination due to fecal intrusion following flooding events and nitrate percolation in groundwater. The present research, indeed, provides, on one hand, new insights into the analysis of the vulnerability in a scant data context; on the other hand, it assesses the adaptation measures contained in the local development plan. Applying the reference definition of the Intergovernmental Panel on Climate Change (IPCC), the results demonstrate that the open shallow water sources in the northern part of the sub-county are more vulnerable to nitrate contamination. Furthermore, the consistency of the results proves the suitability of the methodology selected. Understanding the vulnerability at the local scale is key to planning risk-reduction strategies as well to increasing the local population's knowledge about flood-related risks and water quality.

**Keywords:** vulnerability; rural area; scant data; nitrate contamination; water; flood

#### **1. Introduction**

In Kenyan rural areas belonging to the Arid and Semi-Arid Lands (ASALs), water quantity and water quality are major issues for the local population [1–3]. Although access to water is the main livelihood concern, water quality has an important impact on health. Therefore, the issue of access to clean and safe water has garnered increased attention.

In North Horr Sub-County, in northern Kenya, water quality is threatened, among others, by nitrate contamination after flooding events.

Flood-related hazards, in fact, are posing serious threats to the local population and are worthy of further study. However, flood-related hazards in Kenya are poorly addressed in the literature. In particular, there is a lack of quantitative flood-related risk and vulnerability assessment, mainly due to scant data issues [4], namely the absence or the difficulty of getting access to data in African countries [5–7] as well as problems with data quality (sustainable, continuous, credible, publicly accessible, quality assured dataset) [8]. The most relevant studies are conducted in the framework of vector-borne infectious diseases [9–12] or water-borne diseases, especially fecal-oral diseases [13,14].

This study focuses on nitrate contamination of water in open shallow water sources due to fecal intrusion following flooding events and nitrate percolation in groundwater. This work evaluates the vulnerability of nitrate (NO3 −) contamination in open shallow water sources for human and livestock consumption in a scant data context. Understanding the vulnerability to nitrate contamination is

pivotal since the consumption of contaminated water can have severe outcomes in humans, such as methemoglobinemia, hypertension, increased infant mortality, central nervous system birth defects, diabetes, spontaneous abortions, respiratory tract infections and changes to the immune system [15–19], as well as methemoglobinemia, spontaneous abortion and even death in livestock [20–22].

River and groundwater contamination is generally caused by chemical fertilizer, manure and nitrogenous waste products, all containing nitrogen, used for both agricultural and industrial purposes [23]. Therefore, contamination can affect rural, semi-urban and industrialized areas. In the case here analyzed, the source of nitrates is the deposit of fecal matters. As a consequence of dry climate conditions and pastoralist-related livelihoods, northern Kenya mostly features the availability of open shallow water sources, sometimes found exactly in the riverbed (e.g., open shallow wells) [24]. The seasonal rainfall pattern causes the activation of nitrate contamination hazard. In the case of shallow wells, seasonal streams have an intermittent runoff and pastoralists can let their livestock graze near the well while waiting for watering, allowing the deposit of fecal matter directly into the watercourse. For this reason, there may be direct or indirect contamination of the water source depending on whether "fecal matter is deposited directly into waterways or so close that the potential for wash-in is very high, or via surface runoff and subsurface seepage or drainage" [25]. Although there is certainly a direct pathway of fecal matter in the well during the wet season, it is probable that another means of contamination is nitrate percolation through the permeable surface to the subsurface water reservoir. Nitrate percolation activates during rain events without adequate soil contact time for efficient denitrification and retention [26]. Therefore, soil characteristics modelling is crucial for nitrate vulnerability assessment, but its role in contamination dynamics is not known due to the uniqueness of this means of contamination. Studies in New Zealand focus on nitrate movements along a shallow groundwater flow path in a riparian wetland [26] and on estimating fecal yields from agricultural catchments for water quality purposes [27–29]. Another study in Minnesota assesses private wells vulnerability to nitrate leaching, focusing on future extreme rainfall estimates and floodand nitrate-sensitive well identification [30]. Optimal tools for shallow groundwater risk assessment are the DRASTIC and GOD methods, widely applied in arid contexts [31–35]. The DRASTIC method is considered as one that does not need extensive, site-specific pollution data, but able to provide a solution for evaluating the vulnerability to pollution of groundwater based on known hydrogeological parameters (depth to groundwater, net recharge, aquifer media, soil media, general topography or slope, vadose zone and hydraulic conductivity of the aquifer). The GOD model, indeed, is based on only three parameters (groundwater confinement, overlying strata, and depth to groundwater. However, data availability in a developed context, where the research on nitrate pollution of shallow groundwater is well-developed, is different from a Global South context. In Kenya, these types of data are not available or partially available from project output (difficult to find and obtain) or from a global dataset (based on estimations and not fully reliable). Alternative methods [36,37] require the characteristics of the hazard, i.e., nitrate source yields which are also unknown. In general, therefore, in North Horr Sub-County high spatial and high temporal resolution data, as well as point data, are scarce. Thus, in this paper, a flexible methodology was applied focused on the vulnerability assessment in a scant data context. Since the vulnerability is multidimensional, it is almost impossible to define a universal measurement methodology [38] as well as a finite set of potential indicators [39]. Therefore, using the indications of the Intergovernmental Panel on Climate Change (IPCC) [40,41], nitrate contamination specific sub-indicators of exposure, sensitivity and adaptive capacity were identified and combined in order to obtain the indicator of vulnerability for nitrate contamination hazard in the area. The quantitative approach used in this research to evaluate the vulnerability in the North Horr area is based on a tested methodology [42] applied, for the first time, in a scant data context. Although both qualitative and quantitative methods for the evaluation of the vulnerability from nitrate in open shallow water sources are possible, few studies have tried to tackle the issue with a quantitative approach. The flexible methodology [42] applied aims at linking a quantitative

approach with the socio-economic aspect evaluation [40]. This innovative approach introduces new results about the analysis of the water sources' vulnerability in North Horr Sub-County.

The result of this research may be relevant for other researchers, since this methodology applies to different types of scant data contexts and risks, as well as to decision makers in Marsabit County as an evaluation tool for the measures undertaken to face water issues contained in the development plan.

The remaining part of the paper proceeds as follows. Section 2 gives an overview of the study area and specific key information on the context on which the analysis is based, as well as information on the materials used; Section 3, separated for clarity reasons from Section 2, addresses the methodology of the analysis and the specific issues regarding the construction of sub-indicators of exposure, sensitivity and adaptive capacity. In Section 4, the main results are analyzed and discussed. Section 6 presents the conclusion of this research.

#### **2. Materials**

In a scant data context, like in the northern part of Kenya, availability of data is a pivotal issue. The materials here used have two different types of source: public structured data, i.e., climatic data, demographic data, spatial data, and planning strategies and data collected in the framework of projects. In particular, data on water sources are retrieved from the One Health platform, developed by the Italian start-up TriM (at the moment only for internal use) in the framework of the One Health project (http://www.ccm-italia.org/one-health-uomo-animale-ambiente-north-horr-2) and from the international organization Concern Worldwide (https://www.concern.net/) which shared data collected in a joint campaign with governmental institutions. Except for soil-related data, the data collected were post-processed to construct geo-referenced data in the form of a point vector defined at village or water-source level.

In addition, the limiting factors regarding nitrate behavior, mainly due to scant data context, are identified and assessed. Nitrate behavior in soils, indeed, is dependent on soil characteristics of drainage and texture to consider its transport to groundwater. The role of nitrate degradation was intentionally left out. Attenuation factors like denitrification processes and dilution can occur in the aquifer and influence nitrate concentration [43]. However, the kinetic of nitrate reduction is likely a zero-order reaction and estimations are very low (μM/L per year) [44]. For this reason and for a precautionary hypothesis, denitrification processes are not taken into consideration. Dilution processes are already assessed through the sub-indicators of physical drought exposure and catchment area.

#### *2.1. Study Area: Geographic Positioning and Climatic Context*

The area covered by this study (Figure 1) is North Horr Sub-County (Marsabit County) in the northeastern region with a particular focus on the surrounding area of eight main villages.

This region is considered as part of the ASALs since the area is mostly desert and partially covered by a shrub savannah. Local communities are mostly nomadic pastoralists breeding camels, sheep, and goats with traditional extensive livestock practice. This practice relies only on naturally available resources of pastures and water and requires energy-intensive movements of herds. From the onset of the dry season, pastoralists move to the so-called *fora* pastures and can walk distances up to 40 km to reach the watering points.

Rainfalls have a bimodal pattern (two rainy seasons alternated with two dry seasons) and a high variability of rainfall due to cycles of wet periods and droughts, although the variations of these events are not well known [45–47].

For clarity reasons, we will here refer to the example of shallow wells. However, the existence of other shallow water sources is based on similar characteristics. In this region, due to the bimodal rainfall pattern, watercourses have a seasonal pattern and runoff can appear once or twice a year [24]. During the rainy season, the seasonal stream's flow is restored and the wells in the riverbed are completely inundated. During the dry season, the water supply is provided by wells dug exactly in the streambed. However, the water extracted is not withdrawn from subsurface runoff, but from the

porosity of sand substrates or cones of depression trapped between upward dykes formed of clay soil [24,48]. Those water sources are used both for human and livestock consumption and water can be extracted by hand or by means of motor pumps. For watering purposes, herds of livestock descend the riverbank and wait for their turn, also for hours, close to the well. Therefore, fecal matter can be left directly in the riverbed while waiting for watering. When the stream's flow is restored, nitrate contamination occurs through direct invasion of the fecal matter in the well, added to an indirect component of nitrate percolation in the ground. Thereby, settling water is infected when the water level decreases under the ground level. Then, after the onset of the dry season, water withdrawal and evapotranspiration cause water reduction in the well. Therefore, the concentration of nitrates increases. This process recurs at each season change from wet to dry season.

**Figure 1.** Study area framework. The main figure represents the study area focused on the surroundings of the main eight villages (Dukana, El Hadi, Balesa, El Gadhe, North Horr, Gus, Malabot, Kalacha). The study area is comprised in North Horr Sub-County (Marsabit County) in northern Kenya.

#### *2.2. Climatic Data*

Precipitation data are the result of a quantile mapping bias correction applied to the Kenya Meteorological Department precipitation gridded dataset [49]. Precipitation data cover the study area with high resolution (0.0375 × 0.0375 degrees) for the period 1983–2014 with a decadal temporal resolution.

#### *2.3. Demographic Data*

Population data are derived from the National Census 2019 [50]. However, there may have been underestimation bias due to severe difficulties in taking a census in an area where people are mainly nomadic with very low population density of about 4 persons per square kilometer. The Kenya National Bureau of Statistic itself declared that the enumeration of nomadic pastoralist population is complex and therefore may lead to population underestimation in ASALs.

#### *2.4. Water Sources*

Water sources in the area are mainly represented by boreholes, shallow wells and pans [51], while piped coverage remains limited [52]. Moreover, 66% of sources have contaminated water, which must be treated before drinking [53].

Water resources and allocation data of all kinds are theoretically available for purchase from Kenya governmental agencies, but in reality these data are often difficult or impossible to obtain [54].

Therefore, water source data, the categories of shallow wells, earth pans and rock catchments are retrieved by the One Health platform and through the Concern Worldwide survey. The Second County Integrated Development Plan for Marsabit establishes the presence of 220 shallow wells, 50 pans and 10 rock catchments in the whole North Horr Sub-County. Through the One Health platform and through Concern Worldwide survey, 54 shallow wells, 6 earth pans/dams and 2 rock catchments are identified and analyzed for nitrate contamination vulnerability in the study area (Table 1). Even if shallow wells are the target for the analysis of nitrate contamination vulnerability, earth pans/dams and rock catchments are also taken into consideration as they are subjected to the same conditions of openness, lack of area protection and confinement.


**Table 1.** Water Source Data in the Study Area.

#### *2.5. Hydraulic Conductivity*

The factors that influence the transport and accumulation of nitrate from the land surface to ground water include sediments, rock type and landscape characteristics [43]. However, land characteristics (land use and slope) are here considered negligible factors. The transport capacity of an aquifer is introduced and defined by the fundamental property of hydraulic conductivity (HC) [55]. In fact, the greater the hydraulic conductivity related to the permeability and porosity of the soil, the higher the resulting nitrate infiltration to shallow underground reservoir. The variables taken into consideration are:


#### *2.6. Civil and Hydraulic Structures*

The proposed analysis investigates the vulnerability at water-source level. In order to understand future changes in water source density, attention is drawn to the actions contained in the Marsabit Second County Integrated Development Plan 2018–2022 (SCIDP) [53]. Among the measures that will be undertaken, only new water sources or rehabilitation of existing water sources contained in the SCIDP are taken into consideration.

#### **3. Methods**

This study on spatial vulnerability assessment for nitrate contamination of shallow wells in riverbed is based on the identification of the IPCC approach [40,41]. In particular, the concept of vulnerability refers to "the propensity of exposed elements such as human beings, their livelihoods, and assets to suffer adverse effects when impacted by hazard events" and can be seen as a situation-specific determinant of risk [41].

Therefore, considering the context and the hazard, the model used is [42]:

$$V = \frac{\text{ExS}}{\text{AC}}\tag{1}$$

Both the Exposure (*E*) and the Sensitivity (*S*) represent the negative effects of the changing conditions, while the indicator of Adaptive Capacity (*AC*) is the parameter which may counteract the negative effect of the impact and therefore improves the vulnerability.

Figure 2 presents an overview of the vulnerability assessment model describing the sub-indicators taken into consideration for the construction of this quantitative vulnerability analysis. As previously mentioned, nitrate contamination of open shallow water sources is a climate-driven issue since the wet periods activate contamination pathways and dry periods reduce the contaminant-to-solution ratio (1). The maximum number of settlements, potentially using the water source, is introduces as a two-way sub-indicator considering the exposure of these settlements and the pressure they bring on water sources (2). Moreover, as highlighted by the literature on the topic, nitrate contamination can occur through direct or indirect pathways (3), therefore sub-indicators of lateral and vertical hydraulic conductivity are taken into consideration. The location of the open shallow water sources (inside, outside or near the stream) (4) is introduced as a sensitivity factor while the adaptation strategies are evaluated through the water sources density (5).

**Figure 2.** Vulnerability assessment model. The reference numbers in the red boxes are used in the text to refer to nitrate contamination dynamics in association with the related sub-indicators.

Technical details on the sub-indicators of Exposure, Sensitivity and Adaptive Capacity are provided in Table 2.


**Table 2.** Indicators and Sub-Indicators of the Vulnerability to Nitrate Contamination.

The indicators of Exposure, Sensitivity, Adaptive Capacity and Vulnerability are represented with a pixel-based visualization with a resolution of 230 m (based on the highest resolution represented by the Drainage and Texture sub-indicators' raster resolution), depending on the location of water points, with the aim of providing a local scale vulnerability analysis.

#### *3.1. Exposure*

3.1.1. Physical Exposure: Standardized Precipitation Index (SPI) for Drought and Flood Assessment

The *SPI* [57] is recommended as the main meteorological drought index from the World Meteorological Organization (WMO) [58]. The *SPI* is used to measure the degree of drought and flood stress through a quantitative description method [42,59,60]. The contamination process, indeed, is influenced both by wet and dry periods. The nitrates contained in fecal matter deposited by livestock contaminate the water through direct and indirect pathways that activate in the wet season. The degree of intensity of the dry period influences the water level in the water source, thus, the contaminants-to-solution ratio. Using the SPI, it is possible to detect the susceptibility of the area to wet and dry conditions.

The single parameter required for its calculation is the precipitation (Equation (2)) and it is generally used as an effective and simple tool for drought assessment. The *SPI* index can be calculated for different time scales (from 1 to 24 months) in order to assess different drought types and wetness conditions [61,62]. Short time scales (1 to 3 months) are mainly related to soil water content and river discharge in headwater areas; medium time scales (3 to 12 months) are related to reservoir storages and discharge in the medium course of the rivers; long time scales (12 to 24 months) are related to variations in groundwater storage.

$$SPI = \frac{P - P^\*}{\sigma\_p} \tag{2}$$

where *P* is the monthly precipitation (in mm), *P*<sup>∗</sup> is mean monthly precipitation (in mm) and σ*<sup>p</sup>* is the standard deviation of precipitation.

The *SPI* index was calculated at village level and for the three existing stations present in a 250 km radius (Lodwar, Marsabit and Moyale town) using the *SPI* program (available online at www.drought.unl.edu/droughtmonitoring/SPI/SPIProgram.aspx). The *SPI* index is calculated using the BCKMD dataset, a bias-corrected satellite-derived precipitation dataset based on the KMD dataset (issued by the official national meteorological service dataset, available at http://kmddl.meteo.go.ke: 8081/SOURCES/.KMD/) corrected with the GPCC [49].

The *SPI* index was computed for different time scales depending on the target. For drought detection the 3-month, 6-month and 12-month time scales were computed, while for flood detection the 1-month and 3-month time scales were detected. Specific weight and ranking values were calculated for drought or flood severity and drought or flood frequencies following the method proposed by [63] for drought risk assessment and already applied in risk assessment studies [42]. Using this methodology, the value of the weighted *SPI* can range from a minimum of 0 to a maximum value of 24.

A known limit of the proposed approach is the presence of many zero rainfall accumulations due to the arid climate. However, if there are many historical zero rainfall accumulations, the estimated gamma distribution may not adequately fit the frequency distribution of the historical rainfall. Therefore, in arid regions, the *SPI* indicator should be interpreted with care [64]. Despite this, the use of monthly cumulative precipitation in this research significantly reduces the number of zero rainfall accumulations and the analysis of *SPI* is reliable.

#### 3.1.2. Demographic Exposure

Nitrate contamination affects human and animal populations. However, statistics on animal population are only partially or not at all available. For this reason, since the local population is mainly part-devoted to pastoralism, we can estimate a proportion between human and animal population and we could assume that the greater the human population is, the greater the number of animals would be. However, due to the low population density of the area, the number of settlements was used instead. To understand the sensitivity of the water sources, the demographic exposure was quantified as the potential maximum number of settlements relying on a single water source. In this area, the pastoralists can walk up to 40 km to water their animals. The people in each settlement in a radius of maximum 40 km can potentially reach the water source. However, they would prefer the nearest sources and then gradually move to farther sources. Therefore, the counting of settlements relying on the water source (*Nsettlements*) is weighted using the inverse of the distance between each settlement and the water source (*dx*).

$$N\_{settlemnts=-} = \sum\_{\mathbf{x}=1}^{n} \frac{1}{d\_{\mathbf{x}}} \tag{3}$$

Thus, settlements that are closer to the water source have higher weights, while the farther ones have lower weights. The values of the weighted maximum number of settlements potentially using the water source range from 1.2 (minimum value) to 480 (maximum value).

#### *3.2. Sensitivity*

#### 3.2.1. Presence of the Stream

The presence of the stream constitutes a sensitivity sub-indicator due to the already explained contamination dynamics. In fact, the water points situated in the riverbed are highly sensitive, the water points near the stream are moderately sensitive, the water points and springs far from the riverbed show a low sensitivity (classified as "outside"). The inside/near/outside classification of the different water sources is made through the direct comparison of satellite images and the geo-referenced water sources since this operation would not have been possible with automatic processes. The sensitivity is then expressed through quantitative weights (adapted from [65]) (Table 3).

**Table 3.** Weights Assigned for the Presence of the Stream.


#### 3.2.2. Hydraulic Conductivity

Soil texture and soil drainage classes were reclassified according to coefficients of vertical [66] and lateral [67] hydraulic conductivity respectively (Tables 4 and 5). The soil texture sub-indicator was then averaged across the seven layers to obtain a single vertical HC layer.


**Table 4.** Texture Classes Classifies According to the Coefficients of Vertical Hydraulic Conductivity.

**Table 5.** Drainage Classes Classified According to the Coefficients of Lateral Hydraulic Conductivity.


#### *3.3. Adaptive Capacity*

The evaluation of the adaptation strategies focused on adaptation actions rather than the development or improvement of the institutional framework since they are long-period approaches or other strategies like risk-transfer methods since they are not problem-solving oriented. Other strategies to avoid vulnerability issue like local and conventional actions were discarded. In fact, local strategies in the ASALs against water scarcity mostly rely on traditional adaptation mechanisms, above all, the adoption of nomadic life or increase of watering distance as closer waterpoints are depleted [68]. Conventional adaptation strategies, such as restriction of water use in pans and boreholes during rainy season until surface runoff has been exhausted, controlling the number of livestock that access the pans and boreholes, and paying infrastructure maintenance fees to help increasing water availability for longer periods, are sparsely adopted. However, a more sedentary life and demographic increase are threatening water availability in inhabited areas [69]. For these reasons, the limits of local adaptation capacity must be overcome through the adaptive actions set in place by local governmental institutions. Based on the SCIDP [53], there are eight actions that contribute to the improvement of water sources: installation of gensets and solar panels, construction of shallow wells, supply of fresh and clean piped water, drilling of boreholes, installation of water towers, construction of a dam, installation of underground tanks piped water filled, water trough (Table 6).


**Table 6.** Adaptation Actions (Installation of Gensets and Solar Panels, Construction of Shallow Wells, Supply of Fresh and Clean Piped Water, Drilling of Boreholes, Installation of Water Towers, Construction of a Dams, Installation of Underground Tanks Piped Water Filled, Water Trough) Divided According to the Intervention Areas.

Compared to the current adaptation scenario (AC0), the implementation of the actions stated in the SCIDP will move to a future adaptation scenario (AC1). Therefore, two adaptation scenarios were constructed (AC0 and AC1) where the water sources density in the project area is used as proxy of the sub-indicator of the walking distance to water sources as suggested by [70]. The AC0 scenario, thus, represents water source density ante-SCIDP and the AC1 scenario represents water source density (i.e., ante-SCIDP water sources plus new water sources planned) post-SCIDP.

#### *3.4. Vulnerability*

The vulnerability is the summary indicator obtained from the combination, based on Equation (1), of the previously analyzed indicators and sub-indicators. Each indicator (Exposure, Sensitivity and Adaptive Capacity) was rasterized (WGS 84/UTM zone 37 N, 230 m × 230 m) and normalized in the range 0–10 to be comparable. Then, Exposure and Sensitivity were multiplied together obtaining the potential impacts and finally divided by the Adaptive Capacity counteracting the negative effects of the impacts. As per the scale interpretation, high values correspond to high values of each indicator and of vulnerability.

#### **4. Results**

#### *4.1. Exposure*

#### 4.1.1. Demographic Exposure

The water sources in the surroundings of Dukana show a higher weighted number of settlements and therefore these are the most exposed areas in North Horr Sub-County (see Figure 3a). Some water sources in the western part of the sub-county have a medium demographic pressure. The rest of the water sources are less stressed.

#### 4.1.2. Weighted SPI Index

Although there is a suitable timescale for each specific context, it may be helpful to present results for alternative timescales [71] to observe the variability of the index. SPI index weighted for each main village and station shows different performance according to the time scale (Table 7).


**Table 7.** SPI Index at 3-, 6- and 12-Month Timescales for All the Facilities.

The following considerations on the choice of the timescale for detecting flood and drought spells are since, at this latitude, rainfalls are concentrated into two seasons of three months each. Rain events that trigger fecal matter intrusion in shallow water are extreme events. The 1-month SPI can assess deviation from normal monthly cumulated precipitation better than the 3-month SPI, which can better assess seasonal deviations. Therefore, the 1-month SPI was used as a sub-indicator for physical flood exposure, as it can detect areal susceptibility to wetness conditions (see Figure 3b). The 6-months SPI was preferred to the 3- and 12-month timescale as it can detect two consecutive dry seasons and potentially the onset of a drought period. Therefore, we used it as the sub-indicator for physical drought exposure as it can be very effective in showing reduced streamflow and reservoir levels (see Figure 3c).

#### *4.2. Sensitivity*

#### 4.2.1. Presence of the Stream

The greatest number of water sources that are situated exactly in the riverbed is found in the North and in the East of the sub-county. Therefore, this is where it is possible to find the open water sources that are more sensitive to nitrate contamination (see Figure 4a).

**Figure 4.** Sensitivity sub-indicators: (**a**) presence of the stream, (**b**) soil texture, and (**c**) soil drainage.

#### 4.2.2. Hydraulic Conductivity

The two sub-indicators of hydraulic conductivity (vertical and lateral hydraulic conductivity) describe the permeability to contaminants in the area. In the lowlands in the central-southern part of the North Horr Sub-County there are the greatest rate of vertical hydraulic conductivity due to the presence of the northern part of the Chalbi desert. In the northern part of the sub-county, the vertical hydraulic conductivity decreases with the increase of the altitude (see Figure 4b).

Regarding the lateral hydraulic conductivity, higher values of drainage are found along the watercourses and in the highlands at the starting points of the watersheds (see Figure 4c).

#### *4.3. Adaptive Capacity*

Among the eight actions that contribute to the improvement of water sources, only four contribute significantly to the reduction of the walking distance to safe water sources. These are construction of shallow wells, supply of fresh and clean piped water, drilling of boreholes, installation of underground tanks filled with piped water. Therefore only the following water sources were taken into consideration for the water sources density (Table 8).

**Table 8.** Number and Distribution According to the Type of Water Sources Taken into Consideration for the Water Sources Density in AC0 and AC1.


The increase of the number of safe water sources in the study areas is considered as an improvement of the adaptation capacity. It lowers the animal gatherings at water sources, it reduces the exploitation of water sources themselves and improves access to water.

The scenario AC0, as said, is the current state of density of water sources, while the scenario AC1 is the possible future scenario if all the planned measures are set in place by the end of 2022. In both scenarios, North Horr and El Hadi are the inhabited centers with the higher concentration of water sources.

#### *4.4. Vulnerability*

The vulnerability is the resulting combination of normalised indicators: Normalised Exposure (see Figure 5), Normalised Sensitivity (see Figure 6) combined with the Scenarios AC0 and AC1 (see Figure 7) after their normalization in Normalised AC0 (Figure 8a) and Normalised AC1 (Figure 8b). The vulnerability analysis has a major outcome: both in Scenario 0 and Scenario 1 (Figure 9) the northern part of the sub-county is more vulnerable compared to the rest of area. After the implementation of adaption measures there is a timid improvement in vulnerability values in the southern sub-county (whitish spots). In Scenario 1 (Figure 9b), where there are no new adaptation measures (northern part), the values of vulnerability show no change.

**Figure 5.** Normalised Exposure.

**Figure 6.** Normalised Sensitivity.

**Figure 7.** Adaptation sub-indicator: (**a**) density of water sources (AC0) and (**b**) density of water sources (AC1).

**Figure 8.** *Cont.*

**Figure 8.** Adaptation indicators: (**a**) Normalised AC0 and (**b**) Normalised AC1.

**Figure 9.** Vulnerability indicators in two adaptation scenarios (density of water sources): (**a**) Scenario 0 with the current state of adaptation (**b**) Scenario 1 with implementation of the measures contained in the SCIDP.

#### **5. Discussion**

The result showed that the values of vulnerability in Scenarios 0 and 1 are greater in the northern part of the sub-county (dark red areas) with respect to the southern part (white to light red areas). Very few changes occurred from Scenario 0 to Scenario 1.

In fact, compared to Scenario 0, with the implementation of the adaptation measures—concentrated in the southern and central part of the sub-county—only the areas surrounding Gus, Malabot, North Horr, Kalacha, El Gadhe, Balesa and El-Hadi reduces their vulnerability. The area in the North do not benefit from the adaptation actions. Indeed, the areas in the North that were more vulnerable in Scenario 0 remain the most affected also in Scenario 1, while the areas in the South improve their capacity to adapt to nitrate contamination risk.

The measures contained in the SCIDP that aim to improve water access are concentrated in the southern-central part of the sub-county around the main town of North Horr (Figure 10a). Indeed, it seems possible to observe a radial pattern distribution of the areas of intervention starting from North Horr. As a matter of fact, North Horr is the chief town of the sub-county. Although it is difficult to reach North Horr from Marsabit town passing through the Chalbi desert, the areas in the North are way more difficult to get to and very close to the border with Ethiopia. The organization of similar interventions in the northern area could be challenging from a logistic point of view. Therefore, North Horr seems to be prioritized in the decision-making process. Moreover, the availability of funds is hardly employed in remote areas where the population density is one of the lowest in all of Kenya (about four inhabitants per km2) [50]. In fact, at the moment, funds of sustainable growth are preferably diverted in the areas along the Great North Road (the road going from El Cairo to Cape Town passing through Marsabit County), where a rapid population influx is expected [53].

**Figure 10.** (**a**) Scenario 0 and the planned areas of intervention; (**b**) Scenario 1 and the demographic exposure.

Besides the low population density in the northern part of the sub-county, the demographic pressure on few water sources is an issue. In fact, a great number of settlements rely on few water sources creating significant stress for water supply dynamics (Figure 10b).

Therefore, remoteness of the northern part of the sub-county, security problems, scarce interest, and high demographic pressure on the scarce number of water source create suitable conditions for water scarcity and low water quality. Despite the SCIDP's intention to promote the development of the area for the period 2018–2022, the plan will not be able to rectify the asymmetries and enhance the easiest access to quality water. If not addressed soon, the associated consequences could be an increase in migration sometimes identified as urbanization processes and conflict escalation in the area over water resources. Moreover, worsening consequences could be deteriorated by climate changes and, in particular, by more hydro-climatic variability in space and time [72].

To prevent social inequalities in access to clean and safe water, a broader understanding of the existing water quality in Kenya is necessary. However, poor knowledge and application of the Groundwater Vulnerability Assessment are found [73], even if contamination are common problems that affect groundwater quality [74].

These results and the related discussion confirm that the adopted methodology solves the issue of vulnerability assessment in a scant data context. In fact, the methodology proposed can identify the most vulnerable areas of the sub-county, together with most vulnerable shallow water sources, considering a selection of specific available data. Thus, the scant data limit is overcome and addressed. A limitation could be a lack of groundwater NO3 concentration data; however, an in-situ extensive data collection campaign would be a major effort for the abovementioned logistics issues. Thanks to the methodology proposed, more vulnerable areas are identified and a more narrow group of water sources to be analyzed could be selected, as done by [75]. Alternative applications of this methodology could be useful to assess the vulnerability in other scant data contexts as support for decision making or as preliminary assessment for more in-depth analysis of the water sources regarding nitrate contamination processes.

#### **6. Conclusions**

This research aimed at assessing the vulnerability of the area to nitrate contamination of water in open shallow water sources due to fecal intrusion following flooding events and nitrate percolation in groundwater. The vulnerability indicator was constructed according to two different scenarios pre- and post-interventions contained in the SCIDP, respectively Scenario 0 and Scenario 1. The result obtained shows the partial inefficacity of the interventions since they are concentrated in the central part of the sub-county, where it is already possible to find most of the water sources of the area. The northern part of the sub-county is remote, insecure, and scarcely populated, therefore, there is limited ability and low interest in improving water access in these areas.

Since the adaptation measures planned do not intervene in the most vulnerable areas, at least until 2022, the reduction of nitrate levels in the water supplied may rely on different control methods based on the acceptable level of human health risk, nitrate-control cost and technical feasibility. These methods include nitrate source control, blending of two or more water supplies and direct treatment of nitrates [76]. These methods of nitrates control should be coupled with an awareness campaign on nitrate contamination control.

Moreover, the consistency of the results confirms the goodness of the proposed methodology for scant data context based on the selection of specific quantitative sub-indicators, combining socio-economic aspects with physical features.

In conclusion, the present study contributed to the understanding of the present and near-future vulnerability of the area, as well as of the limitations of the SCIDP in contrasting the risk of nitrate contamination of shallow water sources. Policymakers should be aware of the existing imbalances between the North and South of the sub-county and try to incorporate these findings, albeit achieved in a scant data context, for future policy planning. At the same time, it is demonstrated that a quantitative approach can be easily used to evaluate the vulnerability to climate changes also in a scant data situation and in remote areas.

**Author Contributions:** Conceptualization, V.B. and. A.P.; methodology, V.B., A.P., E.C. and M.R.; formal analysis, V.B.; writing—original draft preparation, V.B.; writing—review and editing A.P., V.B., E.C. and M.R.; supervision, A.P., E.C. and M.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** This study is conducted in the framework of the International Cooperation Project "ONE HEALTH: Multidisciplinary approach to promote the health and resilience of shepherds' communities in North Kenya" funded by the Italian Agency for Development Cooperation (AICS). The authors would like to thank the project coordinator (CCM) and project partners (TriM and VSF-Germany) and the Kenyan Meteorological Department. Finally, a due thank to Concern Worldwide which shared data on water sources collected in a joint campaign with the Government of Kenya.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Mitigation of the Water Crisis in Sub-Saharan Africa: Construction of Delocalized Water Collection and Retention Systems**

**Adolfo F. L. Baratta 1,\*, Laura Calcagnini <sup>1</sup> , Abdoulaye Deyoko 2, Fabrizio Finucci <sup>1</sup> , Antonio Magarò <sup>3</sup> and Massimo Mariani <sup>3</sup>**


**Abstract:** This paper presents the results of a three-year research project aimed at addressing the issue of water shortage and retention/collection in drought-affected rural areas of Sub-Saharan Africa. The project consisted in the design, construction, and the upgrade of existing *barrages* near Kita, the regional capital of Kayes in Mali. The effort was led by the Department of Architecture of Roma Tre University in partnership with the Onlus Gente d'Africa (who handled the on-the-ground logistics), the Department of Architecture of the University of Florence and the École Supérieure d'Ingénierie, d'Architecture et d'Urbanisme of Bamako, Mali. The practical realization of the project was made possible by Romagna Acque Società delle Fonti Ltd., a water utility supplying drinking water in the Emilia-Romagna region (Italy) that provided the financing as well as the operational contribution of AES Architettura Emergenza Sviluppo, a nonprofit association operating in the depressed areas of the world. The completion of the research project resulted in the replenishment of reservoirs and renewed presence of water in the subsoil of the surrounding areas. Several economic activities such as fishing and rice cultivation have spawned from the availability of water. The monitoring of these results is still ongoing; however, it is already possible to assess some critical issues highlighted, especially with the progress of the COVID-19 pandemic in the research areas.

**Keywords:** water crisis in Africa; water collection and retention systems; sand dam; migration; climate change

#### **1. Introduction**

Mali is a landlocked country located in West Africa and 51% of its land is occupied by desert. The cultivated area is 4.7 million hectares, approximately 4% of the entire territory [1]. The country is characterized by the following major types of land:


**Citation:** Baratta, A.F.L.; Calcagnini, L.; Deyoko, A.; Finucci, F.; Magarò, A.; Mariani, M. Mitigation of the Water Crisis in Sub-Saharan Africa: Construction of Delocalized Water Collection and Retention Systems. *Sustainability* **2021**, *13*, 1673. https://doi.org/10.3390/su13041673

Academic Editors: Maurizio Tiepolo, Vieri Tarchiani and Alessandro Pezzoli Received: 28 December 2020 Accepted: 31 January 2021 Published: 4 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

About 47% of the Malian territory is made up of the Niger basin, while the basin of the smaller watercourse called the Senegal River covers 11% of the territory. The Volta basin, the third largest river, corresponds to 1% of the country's surface area, while the remaining 41% is covered by the Sahara Desert. Of note, 1700 km of the 4200 km of the Niger River run through Mali. The Niger and Senegal rivers, and the intricate network of tributaries, provide most of the permanent sources of surface waters. The total average surface water volume is estimated at around 50 km3/year. Niger alone contributes 35 km3/year, a third of which is wasted by evaporation. Renewable water resources, present in the subsoil, can be estimated at around 20 km3/year, half of which is water in common between surface and subsoil. Therefore, in the whole country, the total volume of renewable water is equal to 60 km3/year. The surface water resources entering the country amount to 40 km3/year, mostly from New Guinea (33 km3/year) and from the Ivory Coast (7 km3/year) [5]. For the exploitation of these water resources, the country counts on the presence of five dams (Table 1), for a total water capacity of 13.8 km3 (Figure 1) [6].

**Figure 1.** Water flows of the Niger River system and dams' position (Source: Authors').



Only 5% of renewable water resources available to Mali are exploited. Almost all of the water comes from seasonal surface sources, available only in the period from June to December. Only the withdrawals for the communities come from underground water resources, except for the city of Bamako, whose water is drawn from the Niger River [7]. The reasons why groundwater resources are so poorly exploited can be found in:


Furthermore, the complexity of the country's resource management organization hinders the implementation of any water policy (Table 2).

**Table 2.** Organization chart of the Ministries, Departments and Offices that deal directly or indirectly with water resource management [6].


In addition, a new agency was created in 2002. The *Agence du Bassin du Fleuve Niger* (ABFN) (Niger River Basin Agency) was mandated with safeguarding the Niger basin, as well as the management and integration of water resources in coordination with the corresponding cross-border agencies. The intricate context of governing bodies has given rise to an equally complex regulatory framework through which the management of water resources enjoys little optimization. Ultimately, Mali is characterized by the presence of important water resources, but these are ill distributed over the territory and their strong seasonality is poorly managed.

The access rate to drinking water in Mali is 61% in rural areas and 69.2% in urbanized areas [8]. Bad distribution is followed by bad management, made up of a plethora of bodies with overlapping competencies, scarce powers and a complex regulatory system detached from the local realities (Table 3).

Besides its health implications, access to clean water is considered a human right and is a prerequisite for the realization of other human rights [9]. Therefore, the United Nations, with resolution 64/292 [10], has asked countries and international organizations "to provide financial resources, help capacity-building and technology transfer to help developing countries to provide safe, clean, accessible and affordable drinking water and sanitation for all".

#### **Table 3.** Main regulations on water management in Mali [1].


The paper focuses on the research, design, and implementation process aimed at the construction and upgrade of existing *barrages* in rural areas of Sub-Saharan Africa.

#### *Water Crisis, Climate Change, Internal Conflict, and Migration*

Climate change is among the factors with the greatest impact on the water crisis throughout Sub-Saharan Africa. It will affect those countries, such as Mali, that heavily depend on more less diversified and strongly seasonal agriculture [11].

The climatic conditions of the country (Figure 2) are characterized by average temperatures between 27 and 30 ◦C annually, with large temperature variations occurring mainly in the desertic areas of the north [12]. In 2015, the maximum recorded temperature was 51 ◦C and the minimum was 10 ◦C [13]. The rainy season varies according to latitude: in the south of the country, it lasts up to 6 months, with a marked increase in rainfall between June and October, while in the north it is reduced to just three months between July and September.

The rainfall in the areas close to Sahara is only 50 mm/year, in the Sahel area it is between 100 and 1100 mm/year, while in southern Mali it exceeds 1100 mm/year [12]. Furthermore, Mali is in the so-called Intertropical Convergence Zone where the typical monsoons of West Africa occur. Due to climate change, between 1960 and 2015, the average temperatures increased by 1.2 ◦C, with a future expectation of linear growth: it is estimated that, by 2050, temperatures could increase between 0.9 and 1.5 ◦C, with the largest increase in the Kayes region. These changes could have an impact on the amount

and patterns of rainfall, the main source of hydrological supply. The rain cycle in Mali is decades-long and has undergone a decrease of 4.4 mm/year from 1950 to 1983 and an increase of only 2.6 mm/year between 1983 and 2015 (Figure 3). Mathematical models predict a slight average increase in precipitation (between 1% and 3%) together with a major decrease in the northern driest regions. Furthermore, the variation in the seasonal distribution could shift the wettest period towards the early part, between June and July with a subsequent reduction (between 6% and 10%) for the rest of the period (Figure 3). In addition, rain-related destructive phenomena are expected to increase [14].

**Figure 2.** Mali bioclimatic zones and geographic classification of Kayes region (Source: Authors').

The impact on water resources could be substantial: there would be a reduction in the rate of subsoil resources and, at the same time, an increase in their need because the surface resources would tend to be less available due to intrinsic phenomena (such as the increase in evaporation because of the increase in temperature) and extrinsic (such as the growth in demand for water as a consequence of population growth) [15]. The result of climate change could be a decrease in food security (already under acceptable levels in the regions of Gao, Segou, Tomboctou, and Mopti) and malnutrition (acute childhood malnutrition already affects 13% of children under 5), resulting in increased mortality and reduced life expectancy [16].

The difficult political situation linked to internal conflict, which in June 2019 caused the forced migration of almost 148,000 Malians [17], could exacerbate this perspective. Political instability in Mali has its roots in the period following independence from France, obtained in 1960. The first independent government pursued real-socialism policies and nationalized all the industries in the country except for cotton. In the following five years, the country's economy nearly collapsed and Mali was forced to ask for financial support for currency to its former colonizer. A series of periods of crisis followed, leading to five coup attempts between 1970 and 1990. From the early nineties onwards, multiparty democracy began to consolidate, not without violence. Since 1990, the situation has become complicated

due to the revolts of the Tuareg people, who have settled in the north of the country. They gave life to the revolutionary movement *Mouvement Populaire de l'Azaouad* (MPA) with the aim of "liberating" the territory to the north by force [18]. The conflict officially ended in 1995 and re-exploded in 2007 due to the dissatisfaction of the Tuareg soldiers integrated into the Malian army. In 2012, the MPA was transformed into the National Movement for Liberation of Azawad (MNLA) [19], which found support in Islamic terrorist groups (Al Qaeda and Ansar). Moreover, Libya, until the fall of Gaddafi, supplied arms to the Tuareg that were superior to those available to the Malian army [20]. The declaration of independence of northern Mali, by the MNLA, saw the emergence of conflicting views between the Tuareg and Ansar. The MNLA and the Jihadist group clashed and brought the Tuareg closer to the national government [21]. However, in 2015 the Tuareg accused the Malian government of not respecting the agreements and the terrorist attacks began again.

**Figure 3.** Anomaly of total precipitation [m] compared with anomaly of 2 m temperature (◦C), historical series, in Sahel (12◦ N–17◦ N, 18◦ W–42◦ E). (Data source: Climate Change Institute, University of Maine, 2020).

Climate change, food insecurity, and social, economic, and political instability cause migration. Migration flows abroad come mainly from rural areas (73%), are characterized by a male majority (66%) and have as their primary destination the Ivory Coast (70%). As for long-distance destinations, due to the tightening of entry conditions in some European countries of traditional destinations such as France, the choices of Malian migrants fall on southern European countries such as Spain and Italy.

Rationalizing and improving the management of water resources, especially at the local level, would promote food safety, with the consequent reduction in mortality. In addition, improving the country's agropastoral economy would mean reducing political instability, and, over time, popular revolts. Therefore, a stable water supply would improve the living conditions in the country and reduce migratory phenomena.

#### **2. Materials and Methods**

This three-year research project (2017–2020) started with the cultural and scientific cooperation agreement between the nonprofit organization *Gente d'Africa* and the Department of Architecture of Roma Tre University (P.I., Prof. Adolfo F. L. Baratta, Research unit coordinator Prof. Fabrizio Finucci), with the aim of providing guidelines for the self-construction of infrastructure dedicated to health, food, and water in the depressed areas of Sub-Saharan Africa. Since 2015, the research group of Roma Tre University has been involved with addressing housing, social, and health problems in marginal areas of the world. With the aim of intervening specifically to address the water crisis in Mali, the two partners obtained the operational collaboration of the University of Florence and *Architettura Emergenza Sviluppo* (AES), a nonprofit organization founded in 2016. *Romagna Acque Società delle Fonti* Ltd. provided most of the financing. Furthermore, since 2019, an international cooperation program has been in place between the Department of Architecture of the University of Roma Tre and the *Ecole Superieure d'Ingénierie d'Architecture et d'Urbanisme* (ESIAU) of Bamako (P.I.: Prof. Abdoulaye Deyoko). ESIAU represents the only point of reference for architecture teaching in the entire country. This collaboration produces the planning, design, and construction of small and medium-sized infrastructures, with prevalent hydraulic characteristics in urban and rural areas. These water infrastructures include the distribution network, terminals, and water capture systems aimed at creating artificial basins, the so-called *barrages*. The *barrages* are barriers with a dam function, equipped with locks for the control of flooding.

The project was carried out in the following phases:


The Italian and Malian universities have focused on the opening of the Erasmus Program beyond the European borders, implemented by the European Commission through the Key Action 107, International Credit Mobility, responding to the Call 2019 and being recently (August 2020) awarded the funding for carrying out the activities of cultural cooperation and international educational exchange.

Although the French term *barrages* is mainly used to indicate medium and large dams, in the context of rural settlements it is used to allude to small structures, able to stop or channel the waters coming from the swelling of the streams during the rainy season. In Africa, such structures, also known as sand dams, help the replenishment of the aquifers [22]. *Barrages* are containment structures that can be classified according to the ability to convey subsoil water or surface water into two types (Figure 4):


**Figure 4.** Concept diagram of *barrage* (sand dam): (**a**) Underground *barrages*. (**b**) Surface *barrage* (Source: Authors').

The underground *barrages* are made by digging a deep excavation down to a rocky layer or sufficiently compact soil, allowing for an ordinary foundation of limited size and preventing water from penetrating to greater depths, determining the depth of the aquifer.

The surface *barrages* follow the same principle; however, they are set on a shallow foundation and are designed to stop the watercourse when it swells on the surface because of abundant rainfall on geologically impermeable and compact clay soils.

The *barrage* structure works in a simple but effective way with minimum environmental impact. Because of its small size, it allows overtops water to flow over the *barrage* and avoids draining the aquifer downstream, slowing down the flow of water and facilitating the filling of wells [23].

The surface *barrages* generate an artificial basin which, fed during the rainy season, forms a water reservoir ahead of the dry season. Both *barrage* types, when locally built, use simple construction techniques and intuitive technical principles based on locally available materials. As a result, most of the existing surface *barrages* in Mali are weak structures, often without any foundation. They are incapable of resisting the water pressure as they exploit their weight and shape instead of the characteristics of the material. The poor culture of maintenance means that they are destined to become partially buried due to the accumulation of debris on the upstream-facing side. Furthermore, the lack of a system of local know-how transfer prevents the diffusion of the construction techniques, especially when not based on traditional methods like the use of raw earth, and the correct propagation of a "rule of the art" that could consolidate over the time and could spur any innovations.

However, in many cases the surface and underground *barrages* have been combined in a single structure capable of slowing down the surface flow and increasing the infiltration of water into the subsoil, ensuring greater availability of water resources.

#### **3. Results**

One of the main results of this project was the reconstruction and monitoring of two *barrages*, respectively near the village of Toumbouba and the village of Kofeba, close to Kita, one of the most important towns of the Kayes region, 200 km from Bamako. Both are surface *barrages* although they are different in size and required a completely different approach.

#### *3.1. The Barrage in Toumbouba*

Located at the geographical coordinates 13◦01- 07.58-- North Latitude and 9◦21- 01.04-- West Longitude, the Toumbouba *barrage* (locally known as Tumumba) is grafted onto a preexisting structure built a few decades ago. The original *barrage* was a stone masonry structure, 60 m long, grafted to the north in a retaining wall of an embankment, orthogonal to the position of the main structure. The containing wall resisted the horizontal thrust

due to its shape, having a triangular section with the hypotenuse inclined by about 45◦. The structure was built on ordinary foundation made by an excavation of a few tens of centimeters below the ground level, filled with large aggregate material bound using cement. The elevation structure had a larger size than the foundation, laying at least two thirds directly on the riverbed. The historical memory of the village traces its construction back to the mid-eighties, but the oral tradition did not confirm any certain source.

During the first on-site survey, in April 2017, it was determined that the power of the watercourse in 2007 caused the overturning of the structure and its breaking into three sections. The first section remained firmly buried for its entire length, while the most stressed section, about 15 m long, divided into two parts, laying down on the bed of the watercourse showing the weakness of the construction technique. The torsion due to the overturning caused damage in need of repair to the orthogonal retaining wall, and the erosion of the water caused the dissolution of a large part of the embankment it contained, jeopardizing the necessary shape of the reservoir. During the survey, the research team acquired the preliminary knowledge of the availability of materials and the skills of the local workforce. It learned of the almost absent theoretical and practical knowledge of the use of steel for construction of flexible structures, and, on the other hand, of the recent use of concrete blocks. These could be easily made in large quantities with handmade block molds as demonstrated in the large urbanizations of Bamako and Kita, where they have replaced the use of the traditional Malian raw earth adobe techniques. The survey ended with the inspection of the damaged structure, to proceed with the design phase. The mold for the construction of the concrete blocks for the vertical elevation was designed first, while the two overturned sections were demolished, and the rubble removed. Those operations were carried out with the help of nearby villagers. The design of an armed and buttressed structure was prepared next, one able to offer the greatest resistance to horizontal thrust, combining shape and flexibility. The construction costs were summarized in a bill of quantities, using the prices recorded in the initial inspection phase as a reference and cross-checking them with the international price list integrated in the CYPE software package [24]. The second mission, aimed at the construction, was organized into two phases. During the first phase, in February 2018, the base of the *barrage* was built. While waiting for the setting and curing of the concrete, the concrete blocks were manufactured. During the second phase, in April 2018, the *barrage* was built (Figure 5), and the gaps in the retaining wall of the embankment and on the embankment itself filled.

The result is a wall structure with variable thickness, strongly reinforced in both directions, buttressed, 15 m long and 2 m high, connected to the preexisting structure through the insertion of steel bars Φ16 and concrete injections every 40 cm in elevation for a depth of at least 60 cm. The structure was finished with a 3 cm thick cement-based plaster, to account for the expected lack of maintenance in the future (Figure 6).

During the survey it was decided to intervene with the consolidation of the structure and the reinforcement of the embankment adjacent to the *barrage*. Furthermore, the correct hydraulic functioning of the structure was tested and proved capable of creating a water reservoir of 6.8 hectares with a depth varying between 30 cm and 2 m, corresponding to a water reservoir of about 45 thousand m3 in the period of maximum capacity. The basin serves, directly or indirectly, a very vast territory comprising about 20 villages, or 40 thousand inhabitants.

#### *3.2. The Barrage in Kofeba*

The work for the Kofeba *barrage* benefited from the experience gained by the group during the work in Toumbouba. The introduction of several practical innovations in the workflow resulted in a shortened construction time without compromising full functionality.

Identified by the geographic coordinates 13◦01- 01.77-- North Latitude and 9◦35- 43.60-- West Longitude, the *barrage* is located near a village which in the maps takes the name of Forongo, locally known as Kofeba, located 15 km west of Kita. Like the *barrage* in Toumbouba, the *barrage* in Kofeba was a preexisting structure in stone laid with cement, 215 m long and 80 cm high, with a particular shape whose triangular section provides an acute angle of just 30◦ near the shutter plane. The regulation of the capacity of the basin was carried out through the manual handling of two rudimentary iron closures, hinged as doors without the presence of a linear counter frame or fixed frame.

**Figure 5.** The rebuilding process in Toumbouba: (**a**) the old *barrage* down; (**b**) the building phase; (**c**) the wall complete; (**d**) the *barrage* working phase (Source: Authors').

Over time, this simple moving mechanism has yielded, bending from the rotation plane of the hinges due to the constant horizontal thrust. Because of both the design weakness of this system and the difficulty in finding suitable metal artifacts, it was decided to fill the openings with a wall. Like for the *barrage* in Toumbouba, some demolition works of part of the structure became necessary, because between the two doors there was a wall fragment about 1 m wide.

The demolition left a gap of about 3 m to be filled with the new wall. Given the small size, steps were taken with the aim of making the construction process more efficient and avoid an additional trip. Following a first mission carried out in December 2019, it was possible to ascertain the presence of a solid foundation structure, as wide as the thickness of the base of the wall, or about 2 m, and about 50 cm deep. Therefore, it was decided to proceed with the drilling of holes for the insertion of steel bars ahead of the reinforcement of the structure in elevation, filling the holes with concrete injections. The new wall was built with two imbedded water outlets to control water overflow.

While the work on the Toumbouba *barrage* was carried out without the aid of machinery or construction equipment, for the Kofeba *barrage* it was necessary to use a hammer drill capable of realizing the appropriate depth holes, powered by an electric power generator. In the lapse of time between the two missions, concrete blocks were built with the same procedure already used in Toumbouba, and the material necessary for construction was set aside.

**Figure 6.** The executive design of the *barrage* in Toumbouba (Source: Authors').

The experience previously gained allowed the completion of the construction phase in only 4 working days (Figure 7). The monitoring phase of the Kofeba *barrage* has just started, since, at the time of writing, the rainy season had just begun, but the first photographic surveys to verify the strength of the structure appear satisfactory. At full capacity, in the absence of further settling of the reservoir grounds and with the help of small additional earthworks, the Kofeba basin could cover an area between 9.5 and 20 hectares.

Although belonging to the same typology, the two *barrages* have very different dimensional characteristics (Table 4).

**Table 4.** Comparison between the main data of the two *barrages*.


The Toumbouba *barrage*, although smaller in size, required considerable effort to restore the reinforced concrete foundation along the entire collapsed section. In addition, the water on the Toumbouba *barrage* reaches a very high speed, generating problems of resistance to dynamic thrust. For this reason, both the foundations and the wall in elevations required a greater mass and a longer construction time.

**Figure 7.** The rebuilding process in Kofeba: (**a**) the old *barrage*; (**b**) the building phase; (**c**) the wall complete; (**d**) the *barrage* working phase (Source: Authors').

> Kofeba's intervention (although it is a larger *barrage*) is more contained in terms of size, moreover, it was possible to use the existing foundation section. The Kofeba basin tends to fill gradually with respect to that of Toumbouba and with a lower height of the water level. These aspects have allowed a smaller wall thickness and, therefore, a much shorter period of work. At the same time, however, the Kofeba area required numerous smaller restoration interventions of the natural embankment, carried out with boulders and concrete by the local labor force.

> Both interventions on the *barrage*, for the number of villages and beneficiaries' inhabitants, are characterized by an excellent cost–benefit ratio.

#### **4. Discussion**

The reconstruction of the two small *barrages* has yielded several benefits that can be classified into two categories: direct benefits and indirect benefits.

Regarding direct benefits, the following can be mentioned:


About indirect benefits, the following should be noted:


Given the quality of the results achieved when compared to the very low costs of both interventions (less than 25 thousand euros), the benefits far exceeded the costs incurred. In particular, the benefits can be classified as:


The monitoring phase of the project is still underway to assess its full environmental and economic benefits. A feasibility study is currently looking at the construction of a control station for the acquisition and monitoring of environmental variables, such as temperature, humidity, and pressure, with the aim of demonstrating the micro-climatic variations due to the presence of the water basin.

Data collection relating to the economic benefits indirectly related to the presence of the water basins is ongoing. Several micro-entrepreneurial activities are being observed, such as commercial fish farming, rice cultivation for the sustenance of the community and for retail sale, the appearance of several vegetable gardens, and the introduction of the seasonal rotation of crops.

In particular, in the fish farm in Toumbouba, an artificial pond of about 150 m<sup>3</sup> in capacity was built. This pond was built by excavating; the walls were covered with reinforced concrete, while on the bottom a layer of about 20 cm of compacted earth was created, containing no less than 10% clay, measured by empirical sedimentation tests. This waterproof system, often used for depths not exceeding 3 m, is not recommended for greater depths, where it is advisable to proportionally increase the thickness of the compacted earth. In addition, depths that exceed the height of the aquifer should not be waterproofed in this way to facilitate filtrating of water. The tank requires pumping for the loading and drainage of water, which is periodically necessary for the removal of organic sediments. This periodic maintenance avoids regulating the concentration of oxygen in the water by means of additional mechanical systems. The water of the farm is derived from the *barrage*, by means of mechanical pumping. Therefore, a discontinuous flow was preferred, necessary to replace the evaporated water. To begin breeding, fifty pairs of a

local catfish species weighing between 300 and 400 g were placed in the tank. To facilitate reproduction, artificial nests were inserted into the tank made on the bottom with ballasted polyethylene pipes, 50 cm long with a diameter of 20 cm. Fishes were fed on alternate day manual shedding commercial feed. Despite an insignificant mortality, the fish are sold when they reach a weight between 2 and 2.5 kg, allowing to control the density in the tank. A high density can cause that bigger fish become predator of the others. The small entrepreneurial reality of the fish farm has allowed to create three new jobs in the village.

Favored by a rural economy still largely based on the bartering of goods and services, a series of artisanal activities were born in the villages with the aim of having fish in exchange. Three tailoring and fabric processing activities were born.

The production of concrete blocks for the construction of the *barrage* has allowed the specialization of some inhabitants. Two of them, in particular, have started production (through manual molding) and sale of the blocks to other villages for a total of 10,000–12,000 blocks per year.

Some of these activities are still very fragile and highly dependent on the stability of the retail market. Unfortunately, the recent unpredictable pandemic events have led to a significant reduction in the demand for fish farm products in neighboring markets, leading to the loss of fish stock already in the tanks. The pandemic has highlighted the fragility of entrepreneurial activities undertaken in these contexts that have to face a much higher entrepreneurial risk than elsewhere.

#### **5. Conclusions**

The water crisis in Sub-Saharan African countries, and specifically in Mali, is not a new phenomenon but in recent years has taken on characteristics with geopolitical and international implications. The ongoing climate changes highlight the dramatic consequences of water scarcity: absence of health and hygiene, malnutrition especially in children, with increased mortality rates in under-fives, and the constant decrease in arable land. These factors increase political instability and determine the incessant escalation of revolts, and popular uprisings often linked to international terrorist groups. In these conditions, migratory flows can only intensify, expanding the scope of the phenomenon and establishing the need for international intervention.

To solve the water issue it is necessary to promote the self-construction of small local infrastructures such as *barrages* and to simplify the unnecessarily complex and articulated state management and regulatory systems.

The three-year research project led by the Department of Architecture of Roma Tre University, in collaboration with the University of Florence and in partnership with the Onlus *Gente d'Africa*, *Romagna Acque Società delle Fonti* Ltd. and AES *Architettura Emergenza e Sviluppo*, has led to the reconstruction of two damaged and no longer operating *barrages*, in the Kayes region, near Kita.

The operational phase of the project saw the participation of local communities and the involvement of municipal authorities brought together to restore the functionality and increase the performance of two small and damaged infrastructures.

The research project encompassed a series of on-site missions involving information gathering, surveying, prospecting, design, manufacturing, construction and the collection and dissemination of results. The reconstruction of the basins produced several indirect benefits. These range from improved cohesion of local communities to the increase in microbusiness activities related to the presence of water. In the current phase, the monitoring of the results is mainly aimed at assessing the environmental impact of the *barrages*, through the direct measurement of a series of basic environmental parameters and the indirect detection of the improvement of the living conditions of humans and animals.

**Author Contributions:** All authors wrote the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** The research was supported by Romagna Acque Società delle Fonti Ltd., Gente d'Africa onlus, Roma Tre University, AES Architettura Emergenza Sviluppo.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** A heartfelt thanks to Sara Negroni, Mauro Foli, and Massimo Mantuano from *Gente d'Africa onlus,* And to Tonino Bernabè for *Romagna Acque Società delle Fonti*. A further heartfelt thanks goes to Diawoye Tounkara, local reference for *Gente D'Africa* but, above all, irreplaceable guide, and support in Malian territory. Finally, a special thanks to the students who took part in the various missions in Mali: Giovanni Baratta, Iacopo D'Orazi, Francesca Limongelli and Pietro Marinari.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Article* **Groundwater Resources Assessment for Sustainable Development in South Sudan**

**Manuela Lasagna 1,2 , Sabrina Maria Rita Bonetto 1,2,\* , Laura Debernardi 1, Domenico Antonio De Luca 1,2 , Carlo Semita 1,2 and Chiara Caselle <sup>1</sup>**


Received: 29 May 2020; Accepted: 8 July 2020; Published: 10 July 2020

**Abstract:** The economic activities of South Sudan (East-Central Africa) are predominantly agricultural. However, food insecurity due to low agricultural production, connected with weather conditions and lack of water infrastructure and knowledge, is a huge problem. This study reports the results of a qualitative and quantitative investigation of underground and surface water in the area of Gumbo (east of Juba town) that aims to assure sustainable water management, reducing diseases and mortality and guaranteeing access to irrigation and drinking water. The results of the study demonstrate the peculiarity of surface and groundwater and the critical aspects to take into account for the water use, particularly due to the exceeding of limits suggested by the WHO and national regulation. The outcomes provide a contribution to the scientific overview on lithostratigraphic, hydrochemical and hydrogeological setting of a less-studied area, characterized by sociopolitical instability and water scarcity. This represents a first step for the improvement of water knowledge and management, for sustainable economic development and for social progress in this African region.

**Keywords:** water resources; sustainable management; local development; water for food security

#### **1. Introduction**

Environmental sustainability plays a major role in the United Nations (UN) Agenda 2030 for Sustainable Development. In this framework, the balance between protecting environmental resources and satisfying social and economic needs has become a key development issue.

The global variety of ecosystems and landscapes and the different accessibility to natural resources contribute to shaping social and economic factors, which affect the development of local territories and communities, providing special inputs to their economic growth and resilience. The critical relationship between economic development, growth and environmental protection, if not reconciled in sustainable ways, risks causing the depletion of natural resources.

In South Sudan, as in several Sub-Saharan African countries, environmental conditions are often vulnerable to several shocks (climate, population growth, human activities impact, conflicts and security), which may produce heavy effects on their development.

Many development projects, such as the project "Women Empowerment and Sustainable Agricultural Development to Achieve Food Security in South Sudan (WOSA) AID 10915" funded by the Italian Agency for Cooperation and Development (AICS), support the sustainable management and use of natural resources through actions for protecting water, air and soil, and for preserving biodiversity and combating desertification, including the elaboration and implementation of measures

for mitigating the effects of climate change and for fostering resilience. Other important activities concern the energy and infrastructure sector, which includes transportation and water networks, the distribution of electricity generated by renewable resources and the broad sector of territorial planning.

In developing these approaches, the environmental characterization is promoted to identify landscape character types, geological stratification, biogeographical regions, etc. The recognized spatial frameworks would be useful in assessing interventions and impacts of policies to ensure local food security, social development and sustainable natural resources management. This information should be shared with local administrations, with the scientific community and with any organization that operates in the area.

The economy of South Sudan (East-Central Africa) is predominantly agrarian, with almost 60% of the total workforce engaged either directly or indirectly in agriculture. However, this sector remains underdeveloped due to the political instability of the country and the primitive method of farming systems. Food insecurity is a major problem due to low agricultural production connected with weather conditions and lack of water infrastructure and knowledge [1].

For these reasons, the present study aims to propose an analysis of the local hydrogeological framework in order to answer to the main needs linked to the deficit in the food production and to guarantee a sustainable access to irrigation and drinking water, reducing diseases and mortality.

The study focuses on the city of Gumbo, located at the east side of Juba town. Juba is divided into three sub-districts named Sherikat (Central Gumbo), Jebel Lemon (east of Gumbo) and Adodi (North-West Gumbo). The population of Gumbo is roughly 5000 households, and more than 70% of the population depends on agriculture. About 80% of the population comes from different parts of the country, particularly from east side of Juba. They are internally displaced persons or refugees as a result of the civil war, which broke out in December 2013, and its impact.

The aim of this work is to provide a hydrogeological, lithostratigraphic and hydrochemical reconstruction of the area. These activities are useful for programming drilling campaigns of new wells and in defining sustainable uses or interventions for water according to the quality [2–10]. The shortage of previous studies in the area, also due to the long civil war that occurred in the region in recent years, makes particularly important the collection and the organization of available scientific knowledge on the territory and the execution of new studies to fill the gap of data at the local scale. This may help to create a database for the necessary future interventions that have to be planned in the area.

The study is organized as a first collection of geologic and hydrogeological data of South Sudan, starting from the scarce literature that was completed with a more detailed lithostratigraphic and hydrogeological analysis, specifically focused on the area of Juba.

#### **2. Material and Methods**

#### *2.1. Collection of Information on the Study Area from the Scientific Literature*

The collection of geological and hydrogeological information on South Sudan started from the scarce available scientific literature and brought to the reconstruction of a general framework of the climate, the vegetation and the regional geological and hydrogeological setting. The absence of more detailed data at the local scale was filled with a lithostratigraphic, hydrogeological and hydrochemical reconstruction in the area of Juba, the capital of South Sudan. Specifically, the proposed study focuses on an area surrounding Gumbo (Central Equatoria—Payam Rajaf) (Figure 1), which is located in a flat plain at a distance of about 1.5 km eastwards from the Nile River, referred to here as the "White Nile".

**Figure 1.** Location of the study area (modified from [11]).

#### *2.2. On-Site Survey*

#### 2.2.1. Lithostratigraphic Reconstruction

The lithostratigraphic reconstruction was conducted combining data from a geophysical survey, performed in August 2017 by the United Drilling Company, and from the execution of four new wells drilled in 2018.

The geophysical survey was implemented in an area suffering from lack of water (Figure 2). It consisted of two vertical electrical soundings (VESs), which provided an insight into both thicknesses and types of the overburden strata and of the presence of fractured bedrock, based on resistivity contrasts. VESs were conducted using potential electrode spacing (MN/2) of 0.5 and 5 m, reaching a depth of about 50 m.

#### 2.2.2. Hydrogeological Reconstruction

The new wells were drilled in areas where people lived without free access to water for human, agriculture and livelihood consumption (WPC, W1, W2, W3 in Figure 2). The depth of wells ranges between 60 and 90 m. Stratigraphic logs and water table levels (meters below ground level, m b.g.l.) contribute to the stratigraphic and hydrogeological reconstruction. Water table levels of spring–summer 2018 were interpolated to create a preliminary water table map, indicating the main flow direction of groundwater.

#### 2.2.3. Hydrochemical Analyses

The reconstruction of the hydrochemical setting was performed through two sampling campaigns, conducted in August 2017 (1 water sample, called WPC) and July 2018 (6 water samples). This second campaign includes both groundwater samples (W1, W2 and W3) and surface water samples, taken from seasonal streams (S1, S2 and S3) (Figure 2). Due to the political instability, military checkpoints conditioned the mobility of persons and vehicles, limiting the access to many areas, particularly along

the main streams. Hence, the sampling was only performed for the new drilled wells (groundwater) and areas of free access (seasonal streams samples).

The chemical analyses, performed by the Nuovi Servizi Ambientali (NSA) laboratory in Robassomero (Turin, Italy), included a first step of assessment of data quality, which was accomplished by calculating the balance of positive and negative ions. Water fulfils the principle of electroneutrality and is therefore always uncharged. The level of error was calculated by using the following formula [12]:

• Error of ion balance = ((Σcations − Σ anions) / (Σcations + Σ anions)) \* 100

An error of up to ±5% was considered as tolerable. The results of chemical analyses were then displayed through a Piper diagram.

Since water from wells and seasonal streams is currently used for human consumption, the chemical results were compared with the limits established by the regulations in force in South Sudan and the WHO regulations to provide information on the quality of water in the study area. More specifically, the following regulations were considered:


**Figure 2.** Location of the water samples taken for hydrochemical analyses (wells and seasonal streams) and vertical electrical soundings (VESs).

#### **3. State-of-the-Art Knowledge in the Study Area**

The state-of-the-art geological and hydrogeological knowledge in South Sudan only includes general studies, at the regional or even national scale. The purpose of this review is to collect of all the available information, creating a solid framework on which setting up a more detailed local-scale dataset proposed in this study.

#### *3.1. Climate and Vegetation*

The average altitude of South Sudan is 470 m above sea level, with a gentle slope to the north. In general, the terrain is mainly plain with thick equatorial vegetation and savannah grasslands, and the climate is equatorial. As regards to the annual distribution of precipitation, the rainfall over the River Nile basin is characterized by highly uneven seasonal and spatial distribution. Most of the basins show only one rainy season, typically in the summer months. Only the equatorial zone has two distinct rainy periods. The probability of occurrence and volume of precipitation generally declines moving northwards, with the arid regions in Egypt and the northern region of Sudan receiving insignificant annual rainfall [14]. The area of Juba (specific object of this study) is in a sector characterized by a total rainfall (average annual for the period 1960–1990) between 900 and 1200 millimeters (Figure 3).

**Figure 3.** Average annual rainfall for the period 1960–1990 in South Sudan [14] and location of the study area. The orange line borders the River Nile Basin.

For a detailed description of the study area, data of the weather station in Juba [15] were collected and analyzed. The climate of Juba is tropical with average yearly temperature between 33.8 and 20.7 ◦C. Maximum temperature is during February–March, and minimum temperature is during December–January (Figure 4). The rainy season is from May to October. The dry season starts from November and ends in March and occurs particularly in January and February. Average yearly total rainfall is 974 mm. (Figure 5). The highest number of rainy days occurs from April to October, whereas the lowest number of rainy days are reported in January. The average annual number of rainy days is 104 (Figure 6).

**Figure 4.** Monthly average minimum and maximum temperature of Juba [15].

**Figure 5.** Monthly precipitation of Juba [15].

**Figure 6.** Average monthly rainy days over the year of Juba [15].

Across the Nile region, actual and potential evapotranspiration vary markedly [14]. Seasonal/ monthly variability of evapotranspiration is indeed a function of temperature, wind speed, relative humidity, solar radiation and biomass production. More specifically, the study area is located in a sector characterized by a potential evapotranspiration (average annual for the period 1960–1990) between 1600 and 1800 mL per year (Figure 7).

In the study area, the type of vegetation is biologically diverse, and it is not generally very dense. Examples of common trees are teak, shrubs and mango trees.

The rate of weathering is very high due to the combination of both high temperature and rainfall. This condition facilitates hydrolysis, oxidation and reduction and physical types of weathering such as the action of the plant's roots.

**Figure 7.** Average annual potential evapotranspiration for the period 1960–1990 [14] and location of the study area. The orange line borders the River Nile Basin.

*3.2. Geological Setting*

The geological setting of South Sudan comprises three main geological frameworks [16–18]:


The basement complex is dominated by Proterozoic rocks of medium to high metamorphic grade, with isolated areas of probably older higher-grade granulitic rocks. In broad terms, the Proterozoic metamorphic basement comprises three main units: (i) variably banded and foliated granitic gneiss and migmatites (basement complex, dominantly banded magmatic gneiss in Figure 8), (ii) biotite-amphibole schist/amphibolites and (iii) calcareous meta-sediments and quartz meta-sediments (basement complex, dominantly schists and metasediments in Figure 8).

The crystalline basement complex hosts the majority of mineral occurrences. However, the recorded mineral presence is limited, because the country lacks a significant mineral resource evaluation.

Effusive basic volcanic rocks, mainly basaltic lava and tuffs, are found in the south-eastern part of the country (Rocks volcanics, mainly basalts in Figure 8) and are related to the activity of the East African Rift system.

The Neogene sequence comprises unconsolidated sands, gravels, clay sands and clays (Umm Rwaba Formation in Figure 8) characterized by rapid facies changes. Conditions of deposition of the Umm Rwaba formation are fluvial and lacustrine, with sediments laid down in a series of land deltas similar to those existing in present day South Sudan. The age of sediments is considered to range from Tertiary to Quaternary [19]. This formation hosts most of the ongoing oil exploration activities, and numerous holes/wells were drilled through the cover into the underlying basins. According to oil exploration data from Central and Southern Sudan, the maximum drilled thickness of the Umm Rwaba Formation is higher than 4600 m. According to geophysical data, the maximum recorded thickness is higher than 8200 m in some places [16].

Extensive alluvial deposits underlie the flood plain of the Nile and many of its tributaries within the study area and toward the north of Juba.

**Figure 8.** Map of the geology and mineral deposits of South Sudan [20] and location of the study area.

The Juba area is generally underlain by metamorphic rocks of the basement complex. The basement complex is overlain unconformably by alluvial and surficial deposits that vary in thickness from one location to another. In detail, the basement complex of Juba is composed of gneiss that ranges from medium to high grade metamorphism. Both banded and unbanded gneiss occurs in the area; different types such as augen gneiss, grey gneiss, etc., are present. The gneiss is intruded by several doleritic, gabbroic and granitic intrusions that occur as plutons, or doleritic, gabbroic dikes with east–west trending. The rocks in the study area are mildly deformed. The geological structures consist of both ductile structures, which include folds, foliation and crenulation cleavage, and structures with joints and faults.

Effects of weathering, due to the tropical climate of the area, can be seen in the change of surface color (usually dark or reddish brown) of most rocks, exfoliations and other types of physical weathering.

Close to the Nile River and its tributaries, the geology of the area is dominated by recent alluvium, terraces, deltas and swamp deposits.

#### *3.3. Hydrogeological Setting*

As shown in Figure 9, three different hydrogeological sectors may be identified in South Sudan, based on the structure of aquifers and on the recharge rate [14,21]:


**Figure 9.** Sketch of the structure and recharge rate of the aquifers [14]. The orange line borders the River Nile Basin. The red circle indicates the study area.

The study area is located in a sector with local and shallow aquifers; the groundwater recharge rate is between medium and very low.

The shallow aquifers are usually located in the overburden or in a fractured upper part of the bedrock. The recharge of shallow aquifers is generally dependent on the size of the catchment area and the lithological character of the overburden.

In the study area, the cover sequence may contain a phreatic aquifer if constituted by coarse deposits (gravel, sand, pebble) of a large thickness. Groundwater in the basement formations generally occurs in the weathered (overburden) and fractured rocks. Weathered rocks, indeed, may have a good transmissivity and storage. However, the best aquifers are generally found at the contact zone between the overburden and the rocks. This zone has fewer secondary clay minerals, resulting in a higher transmissivity. Lastly, the highest yielding aquifers can be expected in the fractured bedrock. Boreholes are usually drilled into the fractured bedrock where the permeability is high and where the storage can be provided by the overburden. Fractured aquifers may be recharged through a connected system of fractured zones.

The hydrography of the study area consists of the Nile River and several other streams connected to the Nile River.

#### **4. Results of the On-Site Survey**

#### *4.1. Lithostratigraphic Reconstruction*

The analysis of the stratigraphic logs of the four new drilled wells (Figure 10) highlighted the presence of a clayey and sandy soil on the surface, with a thickness between 1 and 8 m, followed in depth by an alternation of non-fractured granite layers and altered wet granite or fractured granite.

**Figure 10.** Stratigraphic logs of four new wells (not to scale).

The wet granite has a thickness of 2 to 5 m and is located at a depth ranging between 10 and 20 m. The first level of fractured granite has a variable thickness of 2 to 10 m and depth ranging between 18 and 35 m. The second level of fractured granite has thickness of 2 to 6 m and depth ranging between 24 and 65 m.

The screens are mainly positioned in the granitic basement along the entire depth of the wells, with the exception of well W2, which also catches water from the overburden, suggesting the role of both the overburden and bedrock as local aquifers.

The two VESs performed near to the WPC well enabled the identification of layers with homogeneous resistivity (Figure 11). VES1 has slightly lower apparent resistivity values than VES2, but their results are coherent. More specifically, they show the presence in the subsoil of three main layers with different thicknesses and apparent resistivity. In general, the data show the existence of a first layer, with a thickness of about 2 m, characterized by intermediate resistivity that is followed in depth by a second layer with low resistivity, up to about 20–25 m. The resistivity profile ends with a third layer with high resistivity and up to 50 m of depth.

**Figure 11.** Graph with the results of the two VESs and their elaboration.

#### *4.2. Hydrogeological Reconstruction*

Table 1 reports the position, depth and diameter of the wells and the respective water level measurements. The depth of water table ranges between 8.0 and 24.3 m b.g.l. (i.e., piezometric level between 456.0 and 468.7 m a.s.l). The direction of groundwater flow is predominantly SSE–NNW, in the direction of the White Nile, which serves as a drainage axis of the entire surrounding plain sector.

**Table 1.** Data used for the piezometric reconstruction (spring–summer 2018).


Figure 12 reports the resulting water table map. Since the piezometric reconstruction is based on only four wells, the map only indicates the main direction of groundwater flow.

**Figure 12.** Water table map for spring–summer 2018. Due to the scarce data, the map gives indicative information about the hydraulic gradient and groundwater flow direction.

#### *4.3. Hydrochemical Analyses*

Table 2 reports the results of the water chemical analyses. The error of ion balance was always less than ±5%, and consequently it was considered tolerable. Table 2 also reports the comparison with the limits established by the regulations in force in South Sudan and the WHO regulations, providing information on the quality of water for human consumption in the study area.

Results show that both the content in individual elements and the total mineralization are very variable, depending on the analyzed sample. In general, however, surface water from seasonal streams has lower electrolytic conductivity and chloride levels.

The WPC groundwater sample, taken in 2017, shows different chemical features when compared to groundwater samples taken in W1, W2 and W3 wells in 2018. The latter, indeed, have higher heavy metal content. This difference is probably due to the different positions of screens in the wells and thus to the different sampled aquifers.

The limit value of total hardness of the South Sudan regulation (200 mg/L) is only respected in samples W3, W2 and S2, whereas the W1, S1 and S3 exceed it.

The single ions exceeding the suggested limits by the WHO and national regulation that may assume negative effects for human health in the long term and negative consequences for the soil in case of agricultural use are chromium (according to WHO limits) and sodium, fluorides and manganese (with regards to the South Sudan legislation limits).


**Table 2.** Results of chemical analyses and comparison with permissible limits for the South Sudan legislation and WHO guidelines for drinking water quality. The parameters exceeding the permissible limits for the South Sudan legislation are in light blue; those that exceed the WHO guidelines for drinking water quality are in red (N.M. = not measured).


The levels of sodium are usually higher in groundwater than in surface water. WPC, W1, W3 and S3 have sodium concentrations slightly higher than South Sudan limits.

Greater attention must be paid to fluoride, which presents medium–high concentrations (between 800 and 1400 μg/L) in water samples W1, W2, W3, S2 and S3. High fluoride levels may cause damage to bones, kidneys and teeth.

Manganese is usually found in concentrations less than 10 μg/L, with the exception of wells W1 (55.9 μg/L) and W2 (469 μg/L, greater than the South Sudan legislation limit).

The hexavalent form of Chromium, which is very toxic and carcinogenic to humans, is always less than 3 μg/L. On the contrary, the trivalent form, that is not carcinogenic, is present in concentrations higher than the WHO limits for the samples W1, W2 and S3.

To clarify the distribution of parameters exceeding the permissible limits, Figure 13 reports the samples with a content higher than South Sudan national legislation and WHO guidelines, specifying the water concentration.

**Figure 13.** Map of the distribution of the chemical parameters exceeding the permissible limits for the South Sudan national legislation and WHO guidelines for drinking water quality.

#### **5. Discussions**

The set of lithostratigraphic, hydrogeological and hydrochemical data collected with the on-site survey allowed a local framework for the distribution and the quality of the water resource to be defined.

The analysis of VES results and borehole logs coherently suggest the presence of a few meters of overburden (intermediate resistivity) followed by altered/fractured granite (low resistivity) up to 20 m depth and no fractured granite (higher resistivity) at a lower depth. Considering the depth of the wells and the reliability of information coming from direct geological investigation, the borehole logs suggest the alternation of fractured and non-fractured granite up to the end of the wells (60 to 90 m depth).

Based on this lithostratigraphic setting (overburden and igneous bedrock), Figure 14 proposes four possible hydrogeological scenarios. Data suggest the presence of local and shallow aquifers with medium to very low groundwater recharge rate. Phreatic aquifers are hosted in the coarse unconsolidated deposits or in the altered basement, whereas the local aquifers are associated with the fractured basement. The best aquifers are generally located at the contact zone between the overburden and the rocks and in the fractured rocks of the bedrock, which hosts the highest yielding aquifers.

**Figure 14.** Possible scenarios for wells drilled in locations similar to the Gumbo area from a lithostratigraphic–hydrogeological point of view.

This conceptual model is coherent with the hydrogeological framework at the national scale retrieved from the available literature [14,21], which identifies the major aquifers in the northern and in the eastern parts of the state (i.e., on the borders with Sudan and Ethiopia), where the geological setting describes the presence of thick fluvial and lacustrine deposits. The southern and western parts of the region, where the Gumbo area is located, mainly consists of granite and other crystalline rocks, only creating the conditions for local and shallow aquifers. As a consequence of this general assessment, the report [22] describes a difference in the potential between the northern and eastern aquifers and the ones in the south and west wide areas of the country. The former are described as excellent aquifers, with high potential mainly depending on the depth and the good permeability properties of the deposits; the latter, being located in areas characterized by the presence of the basement complex, are described as low potential aquifers, of which yield of groundwater is only enough for rural or urban water supply.

In this hydrogeological framework, despite a general low expected productivity, the local conditions may allow water production to be optimized if the drilling is performed in accordance with the conditions defined as "Productive borehole in fractured zone" (Figure 14), where the boreholes are drilled both in the overburden and fractured bedrock and the latter is characterized by layers of high permeability due to alteration (altered basement complex) or fractures (fractured basement complex). The storage can be provided by the overburden.

The hydrochemical characterization of surface and groundwater is a complete novelty in the panorama of available data on South Sudan. The results furnish the main chemical features of water of the area, normally used for human consumption, irrigation and livelihood.

The data registered a local exceeding of the suggested limits, with assumed possible negative effects for the human health in the long term and negative consequences for the soil in case of agricultural use, for chromium, sodium, fluorides, manganese and total hardness. In general, surface water, coming from seasonal streams, highlighted a lower water quality and scarcer and more ephemeral discharge. Indeed, even if it represents the most direct, cheapest and most easily obtained water resource for humans and animals, it shows higher vulnerability to biological and chemical pollution and is easy to contaminate by anthropic input. Hence, the best supply is represented by groundwater.

Even if the number of samples is very low, some considerations of the relationship between single parameters were made. Chlorides vs. sodium, sodium vs. electrolytic conductivity and chlorides vs. electrolytic conductivity graphs show in general good correlations for surface water (Figure 15).

**Figure 15.** (**a**) Chloride vs. sodium graph (meq/L); (**b**) electrolytic conductivity vs. chlorides graph; (**c**) electrolytic conductivity vs. sodium graph for the surface analyzed waters.

New data have to be added in order to verify the indicative trends obtained.

Chemical data were also plotted in the Piper diagram, allowing for a classification of groundwater and surface water in chemical facies and for a simultaneous comparison of different water samples. The Piper diagram (Figure 16) shows that the samples belong to the Na-HCO3 facies. These waters are typical of deep groundwater environments, influenced by ion exchange.

**Figure 16.** Piper diagram, showing hydrochemical facies of groundwater and surface water samples.

The Na-rich facies resulting from the Piper classification is coherent with the high concentration of sodium that slightly exceeds the national regulation in three groundwater samples out of four. The remaining sample (W2) has, however, an Na concentration of 98.8 mg/L (the limit is 100 mg/L).

Nevertheless, according to the experimental results, this high sodium concentration is not related to an equivalent abundance of chlorides (i.e., the correlation between sodium and chlorides is unbalanced in favor of sodium). The hypothesis of an anthropic contamination that could be considered in presence of salty waters (i.e., similar concentrations of Na and Cl) seems therefore to be unsuitable in this case, suggesting, rather, a natural origin. The described geological framework, reporting a regional bedrock mainly consisting of granite rocks rich in Na minerals (e.g., albite and Na feldspars) supports this hypothesis. Similar results and conclusions were obtained by the author of [23], who proposed an analysis of the water pollution in Sudan (i.e., in the northern geographical position with respect to our study area) and observed a prevalence of sodium bicarbonate water facies where the aquifer was in contact with the rock basement.

A similar natural origin of water pollution may be suggested for the fluorides that exceed the national regulation limits in the majority of the samples. Fluoride was, indeed, described as one of the most important natural pollutants of water in Africa [24] as a consequence of volcanic activities, presence of thermal waters, gases emissions and granitic and gneissic rocks. These natural conditions are often enhanced by low pumping rates, which increase the contact time and favor water–rock interactions, resulting in higher fluoride releases in groundwater [24,25].

As already stated, the concentration of total chromium, which exceeds the WHO limits in three of the samples, is related to a low percentage (<5%) of hexavalent chromium (i.e., the chromium form usually linked to anthropic contamination and more dangerous for the human health). This suggests, again, a natural origin of the contamination connected with lithological setting (i.e., presence of granite or gneiss rocks)

The last anomalous record in the results involves sample W2, which has a content of manganese that slightly exceeds the national regulations and is significantly higher than in all the other samples. Since the depth of the W2 well is not significantly different from the others, this anomalous concentration should be attributed to local conditions. The absence of nitrates and sulphates and the low electrolytic conductivity that characterize this water sample suggest the existence of a reducing environment, possibly connected to an organic contamination that progressively introduces a dissolution of manganese (and iron) already present in the surrounding rocks.

#### **6. Conclusions**

The groundwater resources of a country are crucial for the development of several economic sectors. In semi-arid and arid environments, groundwater enables millions of human beings and animals to exist in terrains that would otherwise be uninhabitable.

The General Assembly of the United Nations published the document "Transforming our world: the 2030 Agenda for Sustainable Development" as a result of the UN summit for the adoption of the post-2015 development agenda. In terms of the Sustainable Development Goal target for drinking water, Sustainable Development Goal (SDG) target 6.1 is to "achieve universal and equitable access to safe and affordable drinking water for all by 2030".

The availability and quality of water, besides representing a fundamental resource as drinking water, is necessary for a good development of agricultural and breeding activities and, consequently, for a sustainable development of the local economy. The water shortage, together with the great frequency and severity of drought periods and the excessive heat conditions, often causes agricultural vulnerability in arid or semi-arid areas of Sub-Saharan Africa. These conditions of vulnerability are enhanced by the global climate changes of recent years, the effects of which are expected to be large and far-reaching predominantly in the developing world [26,27].

For these reasons, the present study proposes an investigation of the lithostratigraphic and hydrogeological setting of the Gumbo village (SE of Juba) to make water extraction systems available and ensure their sustainable management to guarantee food security and economic independence for the populations in the area. Due to the political and social instability, the possibility to move and work in the field during the project was negatively affected, implying the difficulty of collecting data and water samples. Nevertheless, considering the scarce scientific and technical literature available for South Sudan, the collected data and proposed interpretations represent a first fundamental step for the characterization of the Gumbo area. The results contribute to defining a conceptual hydrogeological model and the critical aspects to take into account for the water use in the sampled sites.

In any context, the knowledge of the hydrogeological and hydrochemical setting is an important prerequisite for a risk-informed management of water resources. The planning of hydrogeological-based interventions may guarantee a wide access to good quality water for the local population for drinking, irrigation and livestock purposes. This study represents, therefore, a fundamental step to ensuring food security as well as sustainable economic development and social progress in this African region.

The collected data can be very useful for Administrative Local Authorities, recently established in South Sudan, as well as researchers and the local population, in order to improve water knowledge and management in this little-known area. Moreover, the implementation of a scientific dataset represents the first step toward sustainable economic development and social progress, helping to improve the livelihood of the rural communities and enhancing sustainable and environmentally friendly agriculture, soil conservation and remunerable livestock.

According to the Local Authorities, a survey program increasing the number of water points to monitor the quantitative and qualitative distribution of the water resource should help in better defining the local knowledge gained in the present study and consolidating the scenario that is

defined (higher quality of groundwater than surface water, "productive borehole in fractured zone" hydrogeological model and natural contaminations due to a geological setting with local increasing of exceeding parameters).

Future developments of the study may also include the long-term monitoring of the groundwater. The repetition of the analyses proposed in this study (i.e., measure of groundwater static level, hydrochemical analyses of surface and groundwater samples) at different times throughout the year, together with an increase of the number of sampling points, may allow for the assessment of the real potential of the water resource, with an evaluation of the influence of the seasonal variations on the resource availability. The temporal and spatial increment of collected samples would also contribute to validating the hydrogeological model, with an improved understanding of the flow directions and of the relations between groundwater and surface waters.

Further information should moreover include socioeconomic and demographic data (population, land use, cultivated crops, livestock activities, etc.) of the study area to better understand the various human needs for water to neither exhaust the water sources and the local economy nor have long term negative impact on the environment.

In addition, the monitoring of the chemical contaminations over time, possibly enriched with the local sampling of waters at different depths, may quantify the real danger related to each anomalous concentration and explain the distribution of the contaminations in the aquifer [28,29]. As a function of the specific use (e.g., drinking water, agriculture), the results of these analyses may help to assess the necessity and the features of interventions that may guarantee the safe use of the water resource.

In relation to the intended use of water for agricultural purposes, additional analyses may be specifically directed to the evaluation of the sodium adsorption ratio (SAR), which is a good indicator of the suitability of water for irrigation.

Due to the dry climate conditions of South Sudan, and of the Sub-Saharan African regions in general, another important development of the research could be the assessment of water resources' vulnerability to climate change [30]. An interesting approach to investigate this topic is the analysis of the water resource potential; after the selection of a number of target stations, the study should provide an estimation of rainfall amount on average over the last 30 years and a comparison of rainfall data and survey campaign outputs. The results would show whether the climate change affected the groundwater, with a quantitative evaluation of the entity of this influence on both availability and quality of the water resource [31,32].

A similar development of the research should represent the first step to defining Local Development Territorial Plans that the government needs in order to identify sustainable social and economic assessment (economic activities to encourage, territorial distribution of the refugees, stabilization or development of nomadism). It would be recommended for all decision-makers and politicians but also for rural development program managers to employ a similar approach, which emphasizes the knowledge of the characteristics of the territory and its specificities to deal with the management of natural resources, water in particular, in a sustainable way.

**Author Contributions:** Conceptualization, M.L. and S.M.R.B.; data curation, L.D., C.C. and C.S.; formal analysis, M.L. and S.M.R.B.; funding acquisition, S.M.R.B.; Investigation, L.D.; methodology, M.L. and S.M.R.B.; supervision, D.A.D.L.; visualization, C.C. and C.S.; writing—original draft, M.L., S.M.R.B., C.C. and C.S.; writing—review & editing, M.L. and S.M.R.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by AICS (Agenzia Italiana per la Cooperazione allo Sviluppo), OSC-2016 competition, grant number AID 10915/VIDES/SSD, Title of the project "Women Empowerment and Sustainable Agriculture Development to Achieve Food Security in South Sudan (WOSA)".

**Acknowledgments:** The authors thank Comina Cesare for his help in the interpretation of geophysical data.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Water Resource Management and Sustainability: A Case Study in Faafu Atoll in the Republic of Maldives**

**Maurizio Filippo Acciarri 1,\* , Silvia Checola 2, Paolo Galli 3, Giacomo Magatti <sup>4</sup> and Silvana Stefani <sup>5</sup>**


**Abstract:** This paper contributes to the existing literature in proposing an integrated approach to water management and energy renewable production in a fragile environment. After the 2004 tsunami, in many outer islands in The Republic of Maldives, the lens freshwater natural reservoir was deeply damaged. Currently, the populations of rural atolls use rainwater and water in plastic bottles imported from the mainland for drinking. To provide safe and sustainable drinking water, we analyze the feasibility of two different actions: a desalination system fed by a diesel plant or by a photovoltaic (PV) plant with batteries. The current situation (business as usual, (BAU)) is also evaluated and taken as a benchmark. After illustrating the technical and economic features of desalination and PV plants, a financial and environmental analysis is conducted on the two alternatives plus BAU, showing that the desalination fed by the PV plant results in optimization both on the financial and the environmental side. The levelized cost of water (LCOW) and the CO2 levelized emissions of water (LEOW) are calculated for each alternative. The case study is developed in Magoodhoo Island, Faafu Atoll and can be extended to other islands in The Republic of Maldives and in general to small island developing states (SIDS).

**Keywords:** photovoltaic energy; desalination system; SIDS; CO2 emissions; LCOW; LEOW

#### **1. Introduction**

Most of the small island developing states (SIDS), to which The Republic of Maldives (in short, The Maldives) belongs, share similar sustainable development challenges: a small but increasing population, vulnerability to external shocks, strong dependence on import trade, and a fragile environment [1]. Development is a need and many SIDS now recognize, as a primary objective, the move towards low carbon sustainable economies [1], while at the same time improving the standard of living of the local population.

Moreover, islands enjoy generally beautiful sceneries, fishing resources and unique natural settings that must be preserved and guaranteed for future generations.

As reported by the Asian Development Bank [2] in November 2020, the tourism industry comprises around 25% of the national gross domestic product (GDP) in SIDS. In [3], a road map establishes the guidelines to transition from a fossil-fuel-based energy sector to a cost-effective, business-competitive, affordable, and sustainable renewable energy. This scenario forecasts a continuous and sustained moderate transformation of the energy matrix that will result in 29% of fossil fuels being reduced compared to a business-as-usual situation. As recently reported by IEA [4], the share of renewables in global world electricity generation is nearly 28% in 2020, but there are important differences among countries.

**Citation:** Acciarri, M.F.; Checola, S.; Galli, P.; Magatti, G.; Stefani, S. Water Resource Management and Sustainability: A Case Study in Faafu Atoll in the Republic of Maldives. *Sustainability* **2021**, *13*, 3484. https://doi.org/10.3390/su13063484

Academic Editor: Maurizio Tiepolo

Received: 11 February 2021 Accepted: 15 March 2021 Published: 22 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Even though it is almost universally recognized that renewable energies must gradually substitute fossil fuels [5], the use of renewables on small islands is still quite low based on high transaction costs and poor knowledge of the market potential [6].

Electricity is also dependent on imported fuel, and the level of CO2 emissions is quite high [7]. Taking the Magoodhoo island in The Maldives as a case study, our aim is to analyze an important issue: the coupled nexus water/energy, and to show that feasible solutions are possible, from the financial and environmental point of view.

Scarcity of energy and water are among the major drawbacks for SIDS [8]. This is particularly the case of many outer islands in The Maldives, where, since the 2004 tsunami, the lens freshwater natural reservoir has been damaged, leading to a significant increase in costs as well as an increase in the scarcity of drinking water.

In [9], an extensive review of the recent situation of SIDS is illustrated. Like many SIDS all over the world, The Maldives depends upon imported conventional sources of energy, and scarcity of water severely restricts economic and social development. As far as energy is concerned, the dependence on fossil fuels makes SIDS vulnerable to oil price rises with a significant impact on the local economy. Energy systems and water supply are closely coupled, and the nexus is particularly relevant in SIDS. The ultimate goal is to make SIDS independent from the mainland in energy and water. A desalination system, combined with renewable energy, can definitely be the solution for solving the scarcity of water and contributing to the reduction of pollution from fossil fuels.

In [10], 1087 islands are classified with the use of a cluster analysis according to climatic as well as physical and socio-economic parameters. The conjecture is that an energy supply system proven to be successful on one island could also be successfully implemented on another island.

From the literature analysis, it is quite evident that in SIDS, an energy transition to more sustainable energy sources is desirable and different studies state that the transition should take place in the next few years [11–25].

In The Maldives, energy sources are traditionally based on imported conventional supply, of which the most common fuel is diesel which fuels small power plants, often with low efficiency and heavy polluting. Due to the diesel production, emissions of CO2 into the environment are large and in the medium term may cause damages to the environment. Renewable resources are scarcely used. While solar water heaters are quite widespread, photovoltaic (PV) panels and wind plants are rare. In 2016, the total power generation from renewable energy was 6 MW. In the medium term, the government has planned to install renewable energy systems for up to 30% of the daytime peak load in all inhabited islands [26]. Thus, the efforts of the Maldivian government and local authorities are on implementing renewable systems. The aim is to transit from energy dependence to independence from the mainland.

Studies on energy, applied or theoretical, for the assessment and feasibility of renewable projects in SIDS are quite numerous; less numerous are studies and experiences in desalination combined with renewables.

An extensive study is carried on a reverse osmosis (RO) desalination plant, located in the Sinai Peninsula in Egypt, fed by conventional and unconventional energy sources in [27].

A project for a desalination plant of 1000 m3/day is proposed in the Tyra Island in Greece [28]. The report illustrates two case scenarios: solar and wind power, compared to diesel generators, feed a desalination plant. Diesel generators are used as back up in both cases.

In [29] the feasibility of desalination systems fed by PV or high concentration solar plants of medium and large size in Saudi Arabia is discussed as opposed to diesel as the source of energy. The cost of solar is still high given the heavy subsidizing of diesel in Saudi Arabia.

In Kaya et al. (2019) [30] an analysis based on the levelized cost of water indicates that the current thermal desalination methods used in Abu Dhabi should be substituted by solar PV systems and reverse osmosis (RO) technology for desalination.

Helal [31] explores the economic feasibility of three alternative configurations of an autonomous seawater reverse osmosis (SWRO) unit in remote areas of the UAE. Three different scenarios are proposed here for the powering system where the unit is driven either by a diesel generator, a PV-diesel hybrid system, or solely driven by solar panels without battery backup. This paper can be considered an interesting benchmark, as the low desalination plant dimension [20 m3/day] is comparable with our scenario.

The global average levelized cost of water (LCOW) from desalination plants could decline from around \$2.86/m3 in 2015 to \$1.25 by 2050 if solar, storage systems and other renewable energies are used to decarbonize the sector [32]. The researchers forecast that in specific regions of China, India, Australia, and the U.S., drinking water may cost less than a dollar.

An extensive report from IRENA [33] gives useful information on systems already functioning. Different renewable energy sources are proposed to reduce the environmental impact.

Mainly, studies are based on the use of wind turbines or solar plants, but there are also proposals where other renewable sources, such as sea currents, are used [34]. This application could be interesting particularly for some islands where the scarcity of drinking water is more severe.

A recent paper published in 2020 [35] reports a study about the economic, technical, and environmental impacts of different desalination system configurations (centralized or decentralized, components, and technologies) on transition plans to achieve a higher share of renewable energy and desalination supplies for regions facing water scarcity. The analysis forecasts a reduction of LCOW from the current value of \$2.19–2.46/m<sup>3</sup> to \$0.79–1.01/m3 in 2040.

In The Maldives, scarcity and quality of water are relevant issues. Traditionally, Maldivians used lens freshwater from household wells for both potable and non-potable purposes. The elevation of the highest point in most islands is less than 2 m above sea level, so the freshwater lens thickness of each island is low. The use of rainwater for drinking began in 1978 due to a cholera outbreak. In 1995, the first municipal water supply project began its services. It started from the capital Malé and extended to other population centers, including all touristic resorts. A public water system became available to 49% of the local population and was fully operative by the end of 2017 [36].

However, many outer island populations still depend on collecting rainwater for domestic use, as in the study case illustrated in the present paper. Furthermore, recently and with increased frequency, the collected rainwater stores run out during the dry season, causing expensive emergency water supply from the mainland. Moreover, after the 2004 tsunami, the lens freshwater deposit was deeply damaged, and freshwater is now brackish. While tourist resorts and the city of Malé are all equipped with (thermal) desalination plants, in many outer islands, people use rainwater and a supply of bottled water imported from the mainland for drinking, rainwater for cooking and a small amount of lens water for washing, flushing and other purposes [26].

However, even though The Maldives shows a wide deficit in energy and freshwater resources, they have unique advantages in term of renewable energy (sun) with excellent radiation and seawater potential. An integrated approach to water management and renewable energy production can definitely improve the living in rural atolls and substantially reduce CO2 emissions.

In this paper, we also carried out a life cycle assessment (LCA) for the different solutions proposed. In the literature, different studies on LCA for water systems of different dimensions can be found [37,38]: in integrated urban water systems or for the specific analysis of plastic bottle impact [39].

A synthesis of the literature discussed here, for water desalination and energy source solution adopted, is reported in the discussion section. The data will be used as a comparison with our results.

This paper contributes to the gradual road to energy/water independence from the mainland by proposing a financial and environmental evaluation for the installation and maintenance of a desalination plant. The off-grid energy to the desalination system is provided by diesel or by a PV plant. Results in this paper show that a desalination system plus a PV plant with batteries is the optimal choice, as opposed to the use of diesel, and more feasible than the current scenario, both from the financial and the environmental point of view.

The project is carried out in Magoodhoo Island, in Faafu Atoll in The Maldives. The project is jointly financed by the Marine Research and High Education Center (MaRHE Center). MaRHE Center, officially inaugurated on January 28, 2009, is in Magoodhoo Island. Purpose of the Center is to carry out research and teaching activities in the fields of environmental sciences and marine biology, science of tourism and human geography, to teach how to protect a fragile environment and its biodiversity, how to use and manage its resources in a responsible way. https://marhe.unimib.it/ accessed on 1 March 2021) and by the University Milano Bicocca (Italy) and the Italian Ministry of Environment.

#### **2. Materials and Methods**

The Faafu Atoll is located 137 km from Malé and covers an area of 30 km long and 27 km wide, in the Indian Ocean (Figure 1). The Faafu Atoll is made of 23 islands, covering 1.6 km<sup>2</sup> of land, of which five are inhabited by locals and one is an island-resort in Filitheyo. The capital is Nilandhoo and the population of the atoll is about 4200 inhabitants who live mainly from tourism and fishing.

**Figure 1.** The Faafu Atoll in The Maldives.

Magoodhoo Island is in the Faafu Atoll (3◦4- 26-- N, 72◦57- 15-- E), with a population of 800 inhabitants including the MaRHE Center personnel (35 researchers approx.). In total, there are 133 families of 6 members each on average.

In this project, we have sized a desalination plant on the water needs of the local population. Data collection has been brought forward by two co-authors of the present paper with interviews of local people, inspections of water tanks, and delivery checks of bottled water to the island.

The project involves four main steps: Section 2.1. Understanding the community water demand. Section 2.2. Choice of a desalination technology. Section 2.3. Technological and economic analysis of two alternative actions. Section 2.4. Choice of the best alternative among ACTION 1, ACTION 2 and the current situation (business as usual) based on financial and environmental criteria.

#### *2.1. Understanding the Community Water Demand*

In every family house, including the MaRHE Center, rainwater is used for cooking and drinking, while ground water is used for washing and flushing (see also [36] for a recent survey on the quality of water in The Maldives). In every family house, in the municipality compound and in the MaRHE Center, the ground water is taken from household and public wells, while the rainwater is collected in big plastic water tanks, each containing 2500 L of water. The rainwater pours into the tank through gutters on the roof. The roof is cleaned occasionally, no more than once a year, and the only filters used are pieces of clothes stuffed into the pipe hole. The water tank, once filled up, is closed, and reopened when necessary. The water flows from the tank through a faucet. The tank fills up during the rainy season and occasionally when it rains out of the rainy season. The tanks, provided by the Maldivian government more than 20 years ago and partly by the Red Cross after the 2004 tsunami, are substituted occasionally (Figure 2).

**Figure 2.** Public water tank used by the inhabitants of Magoodhoo to collect rainfall water for daily use.

A reliable estimate puts consumption at 23.81 L per day per family, thereby using up a full tank in approximately three/four months. This amounts to the full use of 3–4 tanks per year (no data about diseases (potential presence of micro-bacteria and plastic microparticles from the tanks) caused by non-filtered water are available). People in the MaRHE Center drink the same water as the village, taken from public tanks, but filter it out by their own filters through a water dispenser located in the MaRHE Center canteen.

The water from the tanks is felt by local people as not safe for drinking, but it is extensively used for cooking. As an alternative to water tanks, water in plastic bottles is used. The bottled water, in plastic usually of 1.5 L, is given mainly to children and sick people to drink and consumption is large. From Table 1, the consumption of water (rainwater and bottled water) per person per day amounts to 4.35 L. The World Health Organization sets the minimum quantity of water needed at 7.5 L per person per day [40]. Statistics on water consumption in touristic resorts and hotels are rare. An estimate from a

big hotel chain in Europe puts water consumption in the range 380–1100 L/guest/night. Swimming pool accounts for an equivalent of 60 L/guest/night [1].


**Table 1.** Water consumption (liters) and cost (\$) per person, per family, of the whole island.

Data in Table 1 correspond to the quantity for drinking and cooking, not considering water for washing and flushing, taken from underground wells.

The bottled water is transported to the village by a supply boat (dhoni) (Figure 3) that carries, on average, 2300 L of water per week. The water is sold in the little shops at an average cost of approximately 3.6 MVR per liter (\$1 = 15.45 MVR, as of March 2019. the water sold in little shops is bought by the local population only. The price of water varies according to supply conditions and demand. People of the MaRHE Center, practicing an anti-plastic policy, drink the water from the dispenser available in the Center. The total consumption of plastic bottled water sums up to 830 L of water per family, per year. This corresponds to an average cost of \$193 per year per resident family the median yearly income of a family in the Faafu Atoll is \$17,087 (264,000 MVR) [http://statisticsmaldives.gov. mv/nbs/wp-content/uploads/2016/03/Presentation-Income-HIES2016.pdf] (accessed on 1 March 2021). The consumption of plastic bottles corresponds to 1.1% of yearly income (Table 1).

**Figure 3.** Weekly supply via dhoni of water bottles for the island of Magoodhoo.

The current individual consumption of water amounts to an average of 4.35 L per day. This water is rainwater and/or bought in the market in plastic bottles.

The aim of the project is to provide 6 L of safe drinking water per day to all inhabitants of the island (Table 2). The supply of water will be provided by a desalination plant fed by an off-grid source of energy that can be renewable or conventional.


**Table 2.** Current consumption and target per person, per family, of the whole island (liters per day).

#### *2.2. Choice of a Desalination Technology*

Given the size of the island, a small scale decentralized desalination unit can be sufficient. The plant will be fed by off-grid electricity.

The strength and weakness of the available desalination technologies were analyzed to choose the appropriate technology. The advantages and disadvantages of some well-known commercial desalination technologies are reported in the literature [41–44].

When choosing a technology, an important factor is the required energy demand for the desalination technology as well as the use of a mature technology when designing a decentralized small-scale desalination unit. In choosing desalination technologies to be coupled with a renewable energy source, it is also important to know the amount of the conventional energy required by the desalination processes (Table 3). The reverse osmosis plant cost has been determined according to an offer by a local installer in Magoodhoo (Table 4). As can be seen from Table 4, the cost of the desalination plant is \$16,500 but the complete installation, including complementary tools, amounts to \$50,000. If working 12 h/day, the plant is designed to satisfy the local demand, guaranteeing 5 m3/day (5000 L/day), which is 40% more of the current consumption.

**Table 3.** Technical specifications of the reverse osmosis (RO) plant.


**Table 4.** Cost of the RO plant.


We recall that a minimum of 7.5 L per capita per day will meet the requirements of most people under most conditions [40]. With this project, the desalination plant will provide 6 L of safe drinkable water per day per person. Rainwater and ground water are still available for other purposes.

The RO plant design includes a pre-treatment process in which fouling elements would be removed. Conventional pre-treatment was selected due to the low incurred costs.

For our requirements, a RO desalination plant of 3.7 kW is selected with a production capacity of 417 L/h. The water produced is stored in two different tanks. As the renewable energy source, photovoltaic is the source commonly coupled with the desalination plants [23,26]. We compare this scenario with a conventional one where the electricity required by the desalination plant is delivered by a diesel generator. Another alternative we consider is the current scenario (business as usual).

#### *2.3. Technological and Economic Analysis of Two Alternative Actions*

RETScreen is a software tool for energy system modelling which analyzes the energy scenario among the energy mix suggested by the user and provides a detailed financial analysis and emission analysis [45,46] software was used to estimate the energy production using solar PV and economic assessment of the plant. The main components of the proposed RO system are shown in Figure 4. The system consists of PV panels, a battery bank, an inverter, a high-pressure pump (HP), a membrane module and a storage tank.

The proposed integrated system for the RO load with 417 kW peak demand and the combined renewable energy sources were modelled using RETScreen software.

First, an energy production analysis involving inputs and costs was carried out for the PV plant. Input values and energy production are illustrated in Table 5. The power capacity of the PV system is 10.8 kW. The output of the inverter in AC kW is 10 kW. Magoodhoo Island, and The Maldives in general, are in a very good position to benefit from the sun's energy through solar PV technology, due to excellent solar radiation all year round. The solar radiation in Magoodhoo varies from 5.9 to 6.9 kWh/m2/day (Figure 5).

**Table 5.** Input values for the RETScreen energy model sheet.


**Figure 5.** Daily average solar radiation in the site [RETScreen data].

The annual electricity produced is about 17,600 kWh, which means a production of about 48 kWh/day. The total solar collector area is 62 m2 and can be located on a roof (see also [24] for a feasibility analysis on roof-mounted solar PV systems in Hulhumale Island in The Maldives for energy production) (Figure 6).

**Figure 6.** Photovoltaic plant under construction on the rooftop of Magodhoo Municipality.

The total average energy consumption is 48 kWh/day to produce 5 m<sup>3</sup> of desalinated water per day (H12 water production).

The cost of the PV plant with all components is \$30,000 (tax included) and has been determined according to an offer by a local installer (Table 6). Prices in the mainland are probably lower, but the island is remote and local installers set the price.



Due to the intermittent nature of solar energy, in a PV based power system, battery storage facilities are needed to ensure a constant power supply. The cost of battery storage is around \$1500. The lifetime is estimated at 10 years.

*2.4. Choice of the Best Alternative among ACTION 1, ACTION 2 and the Current Situation (Business as Usual) Based on Financial and Environmental Criteria*

To proceed with the choice of the most feasible solution, three alternatives have been considered:


For all alternatives, the net present value (NPV), the payback period (PBP) and the CO2 emissions are calculated and compared. NPV is the difference between the present value of cash inflows and the present value of cash outflows over the lifetime of the project. NPV is used in capital budgeting to analyze the profitability of a projected investment. The best alternative has the highest NPV.

PBP refers to the amount of time it takes to recover the cost of the projected investment; or in other words, PBP is the length of time it takes for a projected investment to reach its breakeven point. The best alternative has the lowest PBP.

The current scenario (BAU)

In the current scenario, approximately 6133 bottles of 1.5 L capacity are brought from Malé on the island every month; that is 73,600 plastic bottles per year.

A range between 75 and 180 g of CO2 per bottle of 1.5 L emitted in the life cycle can be estimated [47–52]. This is also confirmed by an LCA evaluation carried out by the Polaris Research Center of the University of Milano–Bicocca. The results state that a 1.5 L plastic bottle is responsible for the emission of 108 grams of CO2 equivalent over its life cycle (data not published). Here, we consider for the LCA of the bottles from the extraction of raw materials to waste disposal.

This means 662.36 kg of CO2 per month and almost 7.95 tons of CO2 per year. We must add to it the emissions produced during the travel by boat. The 1 ton cargo supply dhoni (typology of the boat is referred to the LCA database software Gabi at this link http://gabi-documentation-2018.gabi-software.com/xml-data/processes/3a819ab9 -1979-45b7-a8e9-b4f633d5a662.xml accessed on 1 March 2021) issues in the atmosphere 0.02324 kg CO2 eq per km. The distance from Malé to Magoodhoo Island is almost 137 km, and it is traveled once a week for water transport. It amounts to 25.47 kg of CO2 per month and almost 305.65 kg of CO2 per year.

The average cost of water is 3.6 MVR per liter (\$1 = 15.45 MVR) for a total expenditure of \$25,721 per year (almost \$193 per family per year) (Table 1). This is a cost sensible market change and is expected to grow according to the increasing number of inhabitants. Costs are summarized in Table 7.

**Table 7.** Business as usual (BAU).


intervention), potential damage to health from nonfiltered water for drinking and cooking.

In Table 7, the current scenario is shown. Advantages (pros) are that no investment costs are required. However, disadvantages (cons) are big in damages for the environment and potential dangers for health from the non-filtered water kept in the tanks for months.

ACTION 1 Desalination and Conventional Fuel.

The objective of ACTION 1 is to provide safe water for the community for drinking and cooking.

In this scenario, a RO desalination plant is installed with a working period of H12, 3.7 kW and a capacity of 417 L/h, for a total of 5000 L/day. The market price is \$50,000. In addition, the desalination plant requires an estimated O&M cost of \$1000 per year.

A fuel generator (diesel) powers the plant. The total energy required is 16,200 kWh/y. The fuel generator is sold at a market price of \$5000.

In order to feed the generator (working H12) we need 4524 L of fuel per year. The local price for fuel is 0.91 \$/L which amounts to \$4116.84 per year. This is a yearly cost necessary to power the system, but subjected to market conditions. The emission of CO2/L of diesel is 2.66 kg [21] which amounts to 12,033.84 kg of CO2 issued in the atmosphere per year.

In ACTION 1, the whole demand of water for drinking and cooking is fulfilled. Rainwater is available for other domestic uses, like washing and for personal use. Flushing is guaranteed by the ground water.

From Table 8, plastic bottles (with a saving of CO2 emissions) are not used but the fuel causes high pollution. In case of disruption, rainwater can still be used.

**Table 8.** ACTION 1 Desalination and conventional fuel (\$).


Cons: CO2 emissions are almost 50% more than those in BAU, cost of fuel variable, depending on market conditions.

ACTION 2 Desalination and PV Plant with Battery Storage.

The objective of ACTION 2 is to provide safe water for the community for drinking and cooking.

As in ACTION 1, the market price of the RO desalination plant is \$50,000.

The total energy required is 16,200 kWh/y, provided by the PV plant at 10.8 kW with battery packs of a max capacity of 260 Ah.

In ACTION 2, the initial investment cost is high. The market cost is \$30,000. The price of batteries is \$405 per kWh of installed capacity, which amounts to \$1500 [24] with a life expectancy of 10 years. O&M PV costs amount approximately to \$100 per year.

In ACTION 2, the whole demand of water for drinking and cooking is fulfilled. Rainwater is available for other domestic uses, like washing and for personal use. Flushing is guaranteed by the ground water.

From Table 9, the advantage in terms of the environment is very clear; however, initial capital expenditure is high.

The benefit of investing in ACTION 1 and ACTION 2 comes from the elimination of the "perpetua" costs in BAU by the purchase of water in plastic, which amounts to \$25,721 per year, paid by the whole community.

In ACTION 1, the initial cost of the investment is \$55,000 and the yearly cash flow in the following years is \$20,604.16. This revenue comes from saving \$25,721, obtained by the elimination of plastic bottles, less the fixed cost of fuel to feed the generator (\$4116.84) and the annual desalination plant O&M costs (\$1000).

In ACTION 2, the initial cost of investment is \$80,000 and the future annual benefit is \$24,621, given by the "revenue" \$25,747, which comes from the elimination of plastic bottles, less the fixed costs of maintenance of the PV plant (\$100) and desalination plant (\$1000). The cost of the batteries (\$1500) is added to the cost of the PV plant.

**Table 9.** ACTION 2 Desalination and PV plant with battery storage (\$).


equipped with additional PV modules and battery storage at a higher cost.

Cash flows of the three alternatives, the NPV and the PBP on a 10 year time horizon are calculated. The discount rate (7%) is taken from official 2017 data [53]. Cash flows, NPVs and PBPs are shown in Figure 7 and Table 10 respectively.

**Figure 7.** BAU, ACTION 1, ACTION 2; cash flows.

**Table 10.** Net Present Value (NPV), Payback period (PBP).


From Table 10, it is obvious that action must be taken. BAU at this stage is very expensive for the whole community, let alone the health risks of drinking rainwater kept in plastic tanks for months.

ACTION 1 and ACTION 2 require initial investments that are compensated by the savings from the use of plastic bottles. In a 10 year time horizon, the NPV for both alternatives is high and the PBP is less than three years for both actions. This means that the breakeven point for both actions is reached in less than three years and the initial high investments are compensated by the returns in the next years for the whole community. ACTION 2 shows the best NPV with a slightly higher PBP.

Finally, we calculate the levelized cost of water (LCOW) for all actions, including BAU. LCW allows the collapse of the entire analysis into a unique indicator, which is useful for making the final choice among all available options. LCOW is the cost per m3 of water generated during the whole lifetime. LCW is defined as

$$LCOW\left(\frac{\\$}{\text{m}^3}\right) = \frac{\sum\_{k=0}^n \frac{\mathbb{C}\_k}{(1+i)^k}}{\sum\_{k=1}^n \frac{W\_k}{(1+i)^k}}$$

where *Ck*, *k* = 0, ... , *n* are the yearly cash flows including the initial capital expenditure and *C*<sup>0</sup> . *Wk*, *k* = 1, ... , *n* is the amount of water produced each year. The best alternative shows the lowest LCOW. Using an interest rate of *7*% for a lifetime of 10 years, the results in Table 11. Levelized cost of water (LCOW) and levelized emission of water (LEOW) of each alternative for the whole island. For BAU, we have included rainwater and plastic water, for a total of 1249.944 m3/y. For ACTION 1 and ACTION 2 we have not considered the savings from not purchasing plastic water. For ACTION 1 and ACTION 2, the quantity of water produced has been taken as the maximum capacity of the RO plant (5 m3/day and 1825 m3/y).

**Table 11.** Levelized cost of water (LCOW) and levelized emission of water (LEOW) of each alternative for the whole island.


We have expanded this concept to the levelized emission of water (LEOW) to redirect the impact on the environment in term of CO2 per m3 of water produced in the different alternatives. LEOW is defined as

$$LEOW\left(\frac{\mathbf{g}}{\mathbf{m}^3}\right) = \frac{\sum\_{k=0}^{n} \frac{E\_k}{(1+i)^k}}{\sum\_{k=1}^{n} \frac{W\_k}{(1+i)^k}}$$

where *Ek* , *k* = 0, ... , *n* co are the yearly CO2 emissions and *Wk* , *k* = 1, ... , *n* is the amount of water produced each year.

Results are shown in Table 11. All emissions refer to the production phase: for BAU, that is the use and disposal of plastic bottles; for ACTION 1, the use of diesel after the plant installation; for ACTION 2, there are no CO2 emissions from PV production after the plant installation.

Tables 10 and 11 show clearly that ACTION 2 turns out to be optimal, both in terms of financial and environmental evaluation: ACTION 2 shows the lowest LCOW and no CO2 emissions. On the contrary, ACTION 1 is not sustainable and the impact on the environment does not change practically from BAU.

#### **3. Discussion**

A summary of the literature on desalination plants fed by different energy sources is reported in Table 12. For the sake of comparison, our results are included in the table.

As can be seen from Table 12, the LCOW calculated for actions 1 and 2 (Table 11) is in line with small scale RO plants powered by PV or diesel. Economies of scale apply in desalination plants and LCOW varies according to dimension and location. The production of water on islands, especially if remote, is in general more expensive. Apparently, using only LCOW as an indicator to evaluate the whole project, there is no difference between diesel or PV, as in [30].


**Table 12.** Summary of the literature data; our data are included.

With both ACTIONs 1 and 2 the yearly cost of water per family lowers: while in the current situation it is \$193 (Table 1), with ACTIONs 1 and 2, the cost lowers to \$126 (including a markup of 30% for management of the plants by the municipality). The population pays less for more water, as the desalination plant provides 40% more of the current available water.

However, when considering environmental aspects, the choice is driven strongly towards PV.

Finally, it must be noted that the environmental analysis we have performed is in fact underestimating the effect of plastic disposal since microplastic and charred microplastic may have strongly damaging effect on seawater life. Even though Magoodhoo is remote, the population is relatively scarce and touristic annual afflux is practically null, microplastics and especially charred microplastics are found abundantly nearby the island, related to local practices of burning plastic waste at the shoreline [54]. No indicators exist that can quantify the phenomenon.

#### **4. Conclusions**

Magoodhoo Island, like many rural atolls in The Maldives, suffers from scarcity of energy and water and has a strong dependence on imports from the mainland.

While rainfall is obviously carbon free, the use of plastic bottles has a negative impact on the environment and high costs for the local population.

In order to provide a proper amount of safe drinkable water to the population, installation of a desalination plant is planned, fed by diesel or by a PV plant equipped with battery storage.

Results are in favor of the combination RO + PV plant (ACTION 2), both from a financial and environmental point of view. In less than three years (PBP indicator), the initial investment can be offset by the savings in the use of water imported from mainland. RO + PV is also environmentally friendly (LEOW indicator).

This is one of the first projects in The Maldives for a sustainable water supply system and we strongly believe that if all islands undertake and strive to change the way they manage water and energy significantly, in the next years The Maldives will appreciate a substantial reduction of CO2 emissions, while guaranteeing good water supply.

The local population living conditions must improve, but not at the expense of a fragile environment. Therefore, any choice involving substantial investments must be accompanied by a financial and environmental analysis. Our project considers both aspects.

**Author Contributions:** Conceptualization, S.S. and M.F.A.; methodology, S.S., S.C. and M.F.A.; software, M.A. and G.M.; validation, P.G. and S.C.; resources, S.S. and. S.C.; data curation, G.M.; writing—original draft preparation, S.S. and S.C.; writing—review and editing M.F.A., S.S. and M.F.A.; supervision, P.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** Research developed within the MaRHE Center and B.A.S.E. Center of University Milano Bicocca. The authors are grateful to reviewers for their constructive suggestions.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


#### *Article*

## **E**ff**ects of Land Use Change from Natural Forest to Livestock on Soil C, N and P Dynamics along a Rainfall Gradient in Mexico**

**Daniela Figueroa 1,2 , Patricia Ortega-Fernández 1,3, Thalita F. Abbruzzini 1,4, Anaitzi Rivero-Villlar 1, Francisco Galindo 5, Bruno Chavez-Vergara <sup>4</sup> , Jorge D. Etchevers <sup>6</sup> and Julio Campo 1,\***


Received: 9 September 2020; Accepted: 9 October 2020; Published: 19 October 2020

**Abstract:** The effects of converting native forests to livestock systems on soil C, N and P contents across various climatic zones are not well understood for the tropical region. The goal of this study was to test how soil C, N and P dynamics are affected by the land-use change from natural forests to livestock production systems (extensive pasture and intensive silvopastoral systems) across a rainfall gradient of 1611–711 mm per year in the Mexican tropics. A total of 15 soil-based biogeochemical metrics were measured in samples collected during the dry and rainy seasons in livestock systems and mature forests for land-use and intersite comparisons of the nutrient status. Our results show that land-use change from natural forests to livestock production systems had a negative effect on soil C, N and P contents. In general, soil basal respiration and C-acquiring enzyme activities increased under livestock production systems. Additionally, reduction in mean annual rainfall affected moisture-sensitive biogeochemical processes affecting the C, N and P dynamics. Our findings imply that land-use changes alter soil C, N and P dynamics and contents, with potential negative consequences for the sustainability of livestock production systems in the tropical regions of Mexico investigated.

**Keywords:** climate change; drought; rainfall regime; soil biogeochemistry

#### **1. Introduction**

Globally, soils store at least three times as much carbon (C) as is found in either the atmosphere or living plants [1–3], with the largest portion of it found in tropical forests [4,5]. A global examination of the terrestrial C stocks highlighted that the largest soil organic C stocks in tropical lands are subject to the greatest risks [2], and land-use change continues to pose a threat to tropical forests [6–8]. Moreover, earth-system models have shown that these C stocks in soils will become increasingly vulnerable during the twenty-first century, since reduced rainfall and large-scale agricultural transformation of forests, primarily to pastureland [9,10], are predicted for large areas of the tropics [11]. Tropical

ecosystems dominate the exchange of CO2 between the atmosphere and terrestrial biosphere, yet our understanding of how nutrients control the tropical C dynamics remains far from complete. A better understanding of nitrogen (N) and phosphorus (P) balances in soils can help guide the implementation of mitigation policies and land-use management, since the supplies of one or both nutrients constrain C uptake in the tropics [12]. In contrast to the clear inventory-based assessments of aboveground C on global scales [4], C cycling in soils remains less well understood, due to high biogeochemical variation in the tropics [13].

Environmental factors such as rainfall decline, together with agricultural expansion, generate abrupt, large-scale changes in forestlands and alter soil C, N and P dynamics [14–16]. The impacts of converting natural forests to systems for livestock production on soil C, N and P dynamics have been well examined at the stand scales, with studies relating variations in above- and below-ground C input through litterfall and root exudation to these inherently different management practices [17]. However, the influence of wider regional or global-scale conversion of natural forests to livestock on soil C, N and P dynamics is not yet well understood. This is perceived as a key bottleneck in improving the prediction and evaluation of the results of soil C mitigation efforts related to land-use change; it is also an impediment to better understanding the degradation of soil quality. Previous studies have suggested that converting native forests to livestock systems significantly impacts the quantity and quality of C and nutrient inputs [18,19], and that these changes could be sensitive to rainfall regime. Converting annual cropland to perennial pastureland enhanced microbial biomass and enzyme activities involved in C, N and P cycling due to lower fluctuations in soil water content in pastureland [20]. Campo et al. [21] reported that changes in the land cover can increase surface soil C pools in the dry tropics, while converting native forests to pasturelands in the humid tropics significantly decreases C contents in soils. The changes in land use and management may decrease soil organic matter content and significantly alter soil water dynamics, which would have effects on soil organic carbon [22].

Mexico, with a land area of 1.96 million km2, covers climatic gradients, from humid to dry climate zones, and has experienced diverse and intensive land-use change for animal production, mainly in the tropical regions of the country [23], which are responsible for 50% of national livestock production [24]. Today, beef production in the tropics of Mexico is undergoing major changes with the introduction of management practices for long-term sustainable intensification (intensive silvopastoral systems) through more efficient use of water, materials, and energy [25,26]. Intensive silvopastoral systems are an emerging response to the increasing consumption of animal protein in the country [27] and an alternative to extensive pasture systems, which clear the natural cover to open space for cattle grazing [28,29], reducing the biodiversity and ecological services [30–34], and this intensification of livestock production systems could become a key climate change mitigation technology [35]. For example, intensive silvopastoral systems can produce 12 times more meat than extensive pasture systems [36], and their annual methane emissions per tonne of meat produced are 1.8 times lower than those of extensive pastures systems [37]. With its variety of tropical climates and livestock production systems, Mexico can be viewed as a unique laboratory containing complex interactions between climatic zones and human activities, and thus providing an excellent opportunity to examine simultaneous climate and livestock production impacts on soil fertility, particularly on soil C sequestration and background nutrient status.

To better understand the complex interactions among rainfall, livestock strategies, soil, and C, N and P dynamics, we analyzed C and nutrient dynamics across a gradient of sites varying in rainfall amount (from 1661 to 711 mm year−1), using a robust experimental design. The gradient contains 12 sites, including mature tropical forests (as natural forests), extensive pasture and intensive silvopastoral systems located in humid, subhumid and semiarid climates in the southeast of Mexico. In particular, we studied how C, N and P dynamics in the topsoil layer (upper 10 cm of soil) differed between adjacent native forests and livestock systems, and if those differences were related to rainfall amount or livestock production system (i.e., pastures and silvopastoral). We examined soil bulk organic

C, and total N and P concentrations. We also measured labile concentrations of soil organic C, total N and total P, since their contents cycle more rapidly and should show a greater proportional response to changes in chemistry due to land use, compared to the response of bulk content. For the study, we also include measures of soil basal respiration, net N transformations, and C, N and P enzyme activity, as indicators of soil C and nutrient biogeochemistry.

#### **2. Material and Methods**

#### *2.1. Study Areas*

The study was carried out at four sites with different mean annual rainfall (MAR) in Mexico; two located in a dry tropical region, where MAR decreases from 917 to 711 mm, and two in a humid tropical region, where MAR decreases from 1661 to 1232 mm (Table 1). At each location, we evaluated two types of livestock system: extensive pastures systems and intensive silvopastoral systems. In addition, at each location, a natural forest site (tropical forest in all cases) was used as a reference for original soil fertility conditions. The criteria used for the selection of sites were: (i) sites that shared the same mean annual temperature, but located across a gradient of MAR; (ii) the presence of natural forests, pastures and silvopastures within the same edaphic conditions within each site; (iii) extensive pastures with around 25 years of land use; (iv) silvopastures with around 5 years of land use after around 20 years of extensive pasture systems; (v) both livestock production systems located at a distance of no more than 5 km from the reference (i.e., natural forests). To ensure that none of the sites had been subjected to other human activity, sites were selected following consultations with local owners and a review of the regional government's land-use database.

Long-term climate data from weather stations show that all sites are characterized by a distinct period of low precipitation (four to seven months with rainfall below 100 mm; Table 1). The four sites differ strongly in aridity index (i.e., mean annual rainfall divided by mean annual evapotranspiration). Across the sites, variation in mean annual temperature is less than 1.5 ◦C, and the climate, semiarid to subhumid, would support from tropical dry to humid forests in the Holdridge Life Zone System [38]. Soils (Inceptisols at the wettest end of the rainfall gradient, and Entisols at all three drier sites) have bulk density and clay content that decrease across sites with decreased rainfall amount, and pH that increases from the wettest site to the driest site of the gradient.


**Table 1.** Location and characteristics of four study sites across the southeast of Mexico.

<sup>1</sup> Long-term climatic data (1975–2019 period; *Comisión Nacional de Aguas* personal communication). <sup>2</sup> Aridity Index [39,40]. <sup>3</sup> Water stress months are the number of months per year with rainfall < 100 mm.

Three mature forest stands at each site were selected. The predominant vegetation at the three drier sites is the seasonally dry tropical forest (i.e., forests subjected to prolonged dry season; [41]), while humid tropical forests dominate at the wettest end of the rainfall gradient. Floristically, *Leguminosae* is the most important family across studied sites [42–44], and the abundance of *Leguminosae* trees increases with the decrease in MAR.

The average size of extensive pasture system generally increases with the rainfall amount (87.3 ha in the 1232 mm site, 100 ha in the 711 mm site, 123 ha in the 917 mm and 170 ha in the 1661 mm of MAR site). The extensive pasture system at all sites is dual-purpose, for dairy and meat productions. The average age of pastures of this type was 25 years at the four study sites. At the drier sites (711 and 917 mm of MAR), in pasture systems, one head of cattle per hectare graze for ~2 days and rotate every ~35 days, while an average of 1.3 animals per hectare graze for ~7 days and rotate every ~45 days at the wetter locations (1232 and 1661 mm of MAR). Livestock are usually supplemented with poultry litter and mineral salts at both drier sites, and the pastures do not receive applications of inorganic fertilizers. The most common species used as forage are *Penniseptum purpureum*—Heinrich C.F. Schumacher, *Pannicum maximum*—Jacq., *Cynodon nlemfuensis*—McVaugh, and *Brachiaria brizantha*—Hochst. ex A. Rich. Stapf [45]. Pastures do not receive fertilization at the wetter sites of the rainfall gradient, and the species used as forage are *Brachiaria brizantha*—Hochst. ex A. Rich. Stapf, *Paspalum vaginatum* Swartz, *Cynodon plectostachyus* (K. Schum.) Pilg., *Digitaria decumbens* Stent., *Pennisetum clandestinum* Hochst. ex Chiov, and *Panicum maximum* Jacq. [45]. The establishment of pastures at all study sites was carried out via traditional slash-and-burning of natural forests, followed by ploughing.

The implementation of silvopastoral systems along the rainfall gradient is relatively recent, since the age of the plots range from 4 to 6 years. The establishment of the silvopastoral systems was carried out at sites previously dedicated to extensive pastures for around 20 years, followed by ploughing and the planting of a monoculture of native trees (*Leucaena leucocephala* (Lam.) de Wit.) in high density (~10,000 per hectare). The average animal load in silvopastoral systems is 3 animals per hectare at both drier sites, which graze for ~1 day, while a larger number of livestock (4 to 6 animals per hectare) graze for ~4 days in the wetter sites; cattle are rotated every ~30 days at all four study sites.

#### *2.2. Soil Sampling and Analysis*

Sampling of soils was carried out in the dry (April) and rainy (October) seasons of 2017. One plot (10 × 50 m) was established at each forest stand, pasture and silvopastoral system, and five equidistant (10 m) samples were taken for each type of land use. The topography of all selected plots was even to minimize the effects of local terrain on soil fertility [46]. Samples were taken from the topsoil (0–10 cm in depth), combined into one composite sample per plot and stored at 4 ◦C prior to analysis. A total of three composite samples (from three independent landowners) per site for each type of land use were taken. The upper 10 cm of the soil profile concentrates organic C in tropical dry and tropical humid forests of Mexico [21]. Soil samples were air-dried and sieved (2 mm mesh), and gravimetric water content in fresh soil samples was determined prior to analysis to correct the soil weight used in each determination. Although water content in fresh soils did not vary among land uses or across sites, samples taken in the dry season had less moisture than those taken in the rainy season (23.7 ± 2.1 and 43.9 ± 4.1%, respectively; mean ± 1 standard error).

Soil texture [47] and pH (in water) were determined prior to C, N and P analyses. Total and inorganic C (carbonates) concentrations were analyzed in an automated C analyzer (SCHIMADZU 5005A), by grinding a 5-g air-dried subsample (100-mesh screen). Organic C was estimated from the difference between total and inorganic C concentrations. The concentration of total N and total P was determined by acid digestion [48] using an NP analyzer (Technicon Autoanalyzer III). Soil available P was extracted using the Bray and Kurtz [49] method for acidic soil (pH ≤ 7), and the Olsen method was used for alkaline soils (pH > 7). After extraction, available soil P concentrations were determined with the colorimetrical method [48].

Carbon and N concentrations in soil microbial biomass were determined by chloroform fumigation– extraction methods [50] using replicated samples of fresh soil. Fumigated and non-fumigated samples were incubated at 24 ◦C for 24 h. Microbial biomass C was extracted with 0.5 MK2SO4, filtered (Wathman No. 42 paper), and the concentration of C was measured using an automated C analyzer. Microbial C was estimated from the difference between C concentrations in the fumigated and the non-fumigated extracts, and a conversion factor *kC* equal to 0.45 [50] was used. Microbial biomass N was extracted in a similar way, after filtering through Wathman No. 1 paper, the filtrate was digested in acid and the total concentration of N was determined using an automated NP analyzer. Microbial N concentration was calculated in a similar way to microbial C, using a conversion factor *kN* of 0.57 [51].

Soil basal respiration was determined from duplicated fresh subsamples (20 g) incubated at 25 ◦C. Soil subsamples were moistened to 50% water-filled pore space following light tamping in a jar (700 mL) containing vials of water to maintain humidity and 10 mL of 1.0 M NaOH to absorb CO2. Alkali traps were replaced at 1, 2, 3, 5, 7, 14, 21 and 28 days and were removed at 35 days. Evolved CO2 was determined by titration of alkali with 0.5 M HCl [52]. Soil basal respiration was calculated using measurements from days 3 to 35 to avoid the majority of flush due to drying and rewetting. The CO2 flux produced was standardized per gram of soil organic C.

We measured mineral N concentrations (NO3 plus NH4) and net N mineralization and nitrification rates using 2 M KCl extraction and aerobic incubation methods. Mineral N concentrations were measured by extracting a 15-g sub-sample in 100 mL 2 M KCl [53]. The soil KCl solution was shaken for 1 h and allowed to settle overnight. A 20 mL aliquot supernatant was transferred to vials and frozen for analysis (initial mineral N concentration). Nitrogen mineralization and nitrification rates were measured during 15-day aerobic incubations [53]. A second sub-sample was wetted to field water holding capacity with distilled water, maintained at field capacity moisture and incubated at 25 ◦C for 15 days before extraction, using KCl (final mineral N concentration). Analysis of both initial and final mineral N concentrations was done on an Autoanalyzer system using procedures to determine NO3–N plus NO2–N, which were reported as NO3–N, and using the salycilate–hypochlorite procedure for NH4–N. Nitrogen mineralization rate was determined from the difference between mineral N concentrations at the start and end of the incubation, and results were expressed on a basis of mean daily mineral N production. Likewise, nitrification rate was determined from the difference in NO3–N concentration at the beginning and end of the incubation, and results were expressed in similar units.

We measured the indicator enzymes most commonly used to infer growth-limiting C sources and nutrients: ß-1, 4-glucosidase (BG; enzyme commission number: EC 3.2.1.21) and polyphenol oxidase (POX; EC 1.10.3.1) to infer C-acquiring enzymes, ß-1, 4-N-acetylglucosaminidase (NAG; EC 3.2.1.14) to infer N acquiring enzymes, and acid phosphatase (AP; EC 3.1.3.1) to infer P acquiring enzymes [54], following the method proposed by Jackson et al. [55]. The enzyme assays were incubated at 25 ◦C for 2 h and their absorbance was recorded using a microplate reader at 410 nm for ß-1, 4-glucosidase, ß-1, 4-N-acetylglucosaminidase and acid phosphatase activities, and at 460 nm for polyphenol oxidase activity. The concentration of pNP (or tyrosine, for polyphenol oxidase) detected in the soil sample was corrected by subtracting the sum of absorption from the sample and substrate control wells, and enzyme activities were calculated as follows:

#### *EA* = (final absorbance)/(*C* × incubation time × soil dry mass)

where *EA* is the enzyme activity expressed in μmol of pNP (or tyrosine, for polyphenol oxidase) released per gram of soil per hour (μmol g−<sup>1</sup> h<sup>−</sup>1), and *C* is the conversion factor that relates absorbance to μmol of pNP (or tyrosine, for polyphenol oxidase) for each enzyme activity.

#### *2.3. Statistical Analyses*

For each metric, a one-way ANOVA was performed, testing the effects of the rainfall regimen and the effect of the land-use change. The ANOVA residuals were explored for normality using the Shapiro–Wilk's test. Data were transformed logarithmically when the assumptions of normality did not occur; the following soil metrics violated the normality assumptions: microbial biomass C and N concentrations, net N transformation rates and the activity of the ß-1, 4-glucosidase. The honest significant difference (HSD) test was used when statistical differences (*p* < 0.05) were observed across sites, or among land uses in each site. The interactions between soil metrics, land uses (natural forests, pastures and silvopastoral systems) and rainfall regime (MAR) were explored using a principal component analysis (PCA). In addition, correlation matrices were used as a way to depict the relationships between soil metrics within each site in both sampling seasons (i.e., dry and rainy seasons). All statistical analyses were performed using R statistical software [56].

#### **3. Results**

#### *3.1. Carbon, Nitrogen and Phosphorus Concentrations*

Soil organic C, total N, total P and, microbial biomass C and N concentrations were the highest at the driest end of the rainfall gradient and decrease with rainfall increase, meanwhile NH4 concentration was higher in the wettest end of the rainfall gradient (Figures 1–3, Table 2). Rainfall regime did not have a significant effect on soil NO3 and available P concentrations (Figure 2). Rainfall seasonality did not affect the organic C, total N, total P, and microbial biomass C and N concentrations in soils. However, NH4 and NO3 concentrations were higher in the rainy season than in the dry season. In contrast to the seasonal variation in soil inorganic N concentrations, available P concentration was consistently higher in soils taken in the dry season than those taken in the rainy season.

Land-use change from natural forests to both livestock production systems (i.e., extensive pasture and intensive silvopastoral systems) decreased soil organic C, total N and P available concentrations (Figures 1 and 2, Table 2). In contrast, no significant changes in NO3, NH4, total P and microbial biomass C and N concentrations were detected with land-use change (Figures 1–3). Finally, no significant changes in soil C, N and P concentrations were detected between livestock production systems.


**Table 2.** Site, season and land-use effects on soil carbon, nitrogen and phosphorus metrics along a rainfall gradient in Southeast Mexico.

Significance main effect: NS, *p* > 0.05; \*, *p* < 0.05; \*\*, *p* < 0.01; \*\*\*, *p* < 0.001.

**Figure 1.** Soil organic carbon, total nitrogen and total phosphorus concentrations in dry (**a**–**c**) and rainy (**d**–**f**) seasons under natural forests, pastures and silvopastoral systems along a rainfall gradient. Data are means and confidence intervals.

**Figure 2.** *Cont*.

**Figure 2.** Soil ammonium, nitrate and available phosphorus concentrations in dry (**a**–**c**) and rainy (**d**–**f**) seasons under natural forests, pastures and silvopastoral systems along a rainfall gradient. Data are means and confidence intervals.

**Figure 3.** Soil microbial biomass carbon and microbial biomass nitrogen concentrations in dry (**a**,**b**) and rainy (**c**,**d**) seasons under natural forests, pastures and silvopastoral systems along a rainfall gradient. Data are means and confidence intervals.

#### *3.2. Soil Basal Respiration and Net Nitrogen Transformations*

The soil basal respiration was greater in wetter sites (i.e., sites with 1232 and 1661 mm of MAR) than in the drier counterparts (i.e., sites that receive 711 and 917 mm of MAR) (Figure 4, Table 2). In addition, net N transformations differed considerably among sites reflecting changes in rainfall amount. However, the paired comparisons using the Tukey–Kramer HSD test show that soils from the driest end of the rainfall gradient had the highest N transformations, and net N mineralization and net nitrification rates decreased with increase in rainfall amount. The season did not have a significant effect on soil basal respiration. However, large variation in net N transformations between seasons were observed, with higher net N mineralization and net nitrification rates in the rainy season than in the dry season.

Land-use change from natural forests to extensive pasture and intensive silvopastoral systems increased soil basal respiration (Figure 4, Table 2). However, net N transformations did not change with land-use change irrespective of the livestock production system. No significant changes in these soil C and N fluxes were observed between pasture and silvopastoral systems.

**Figure 4.** Soil basal respiration, net nitrogen mineralization and net nitrification in dry (**a**–**c**) and rainy (**d**–**f**) seasons under natural forests, pastures and silvopastoral systems along a rainfall gradient. Data are means and confidence intervals.

#### *3.3. Enzyme Activities*

The activity of C- and N-acquiring enzymes varied significantly among sites; but changes were not consistently related to rainfall regime (Figure 5, Table 2). Intermediate sites across the gradient (i.e., sites that receive 911 and 1232 mm of MAR) had higher polyphenol oxidase and 4-N-acetylglucosaminidase, or ß-1, 4-glucosidase activities, respectively, whereas the wettest site showed the lowest activity for the three enzymes. In contrast, soils from the wettest end of rainfall gradient had the highest acid phosphatase activity, reflecting the low concentration of available P. Rainfall seasonality did not have consistent effect on the activity of C-, N- and P-acquiring enzymes (e.g., on polyphenol oxidase, ß-1, 4-N-acetylglucosaminidase, and acid phosphatase) across sites. However, generally the activity of ß-1, 4-glucosidase was greater in soils collected in the dry season than in soils collected in the rainy season.

The land use did not affect the activity of C-acquiring enzymes in soils (Figure 5, Table 2). However, there were significant differences in the activity of N- and P-acquiring enzymes between natural forests and intensive silvopastoral systems across sites, with an increase in the ß-1, 4-N-acetylglucosaminidase and acid phosphatase activities under the legume-silvopastoral system. No significant changes in Nand P-acquiring enzyme activities were detected in between extensive pastures and natural forests, or with intensive silvopastoral systems.

**Figure 5.** *Cont*.

**Figure 5.** Activities of ß-1, 4-glucosidase, polyphenol oxidase, ß-1, 4-N-acetylglucosaminidase and acid phosphatase in soils in dry (**a**–**d**) and rainy (**e**–**h**) seasons from natural forests, pastures and silvopastoral systems along a rainfall gradient. Data are means and confidence intervals.

#### *3.4. Multivariate Analysis and Soil Metric Relationships*

The PCA allowed us to observe the effects of rainfall on soil metrics in each land-use system in the dry and rainy seasons (Figure 6). The two firsts principal components explained 61 and 52 percent of data variation for the dry and rainy seasons, respectively (Tables 3 and 4). Across the entire data set, the first principal axis summarized 47 percent of the variation in soil metrics in the dry season (Figure 6a, Table 3). Organic C, total N, NH4 and acid phosphatase had the highest correlation scores on this axis. Sites found at the negative end of PC1 exhibited the lowest activity of acid phosphatase (pastures and silvopastoral systems at 1232 and 1661 mm of MAR) and the highest NH4 concentrations (pastures and silvopastoral systems at the wettest end of the rainfall gradient) (Figures 2 and 5). By contrast, sites found at the positive end (natural forests and pastures at the driest end of the rainfall gradient) exhibited the highest organic C and total N concentrations (Figure 1). The second axis accounted for 14 percent of the variation. Available P and polyphenol oxidase explained the second-largest fraction of the explained variation in soil characteristics. Sites found at the negative end of PC2 exhibited the highest polyphenol oxidase activities irrespective of land use (sites at 917 mm of MAR). By contrast, sites found at the positive end (all land uses at the driest end of the gradient) exhibited low phosphatase activity.

On the other hand, in the rainy season, the first principal axis summarized 34 percent of the variation (Figure 6b, Table 4). The importance of organic C, total N and acid phosphatase detected in the soils of the dry season was also observed in the first PC of the rainy season soil samples. Sites found at the negative end of PC1 exhibited high values in acid phosphatase activities under livestock production systems (the wetter sites; i.e., extensive pastures and silvopastoral systems at 1232 and 1661 mm of MAR). Furthermore, the positive end of the PC1, reflect the highest organic C and total N concentrations in soils under the natural forests at the 711 mm of MAR. The second

axis accounted for 17 percent of the variation. Polyphenol oxidase and microbial biomass C and N explained the second-largest fraction of the explained variation in soils. Sites found at the negative end of PC2 exhibited the highest concentrations of microbial biomass C and N under extensive pastures and intensive silvopastoral systems at the driest end of the rainfall gradient. By contrast, sites found at the positive end of this spectrum exhibited the highest polyphenol oxidase activity under extensive pastures and intensive silvopastoral systems, but in sites at 917 mm of MAR.

**Figure 6.** Principal components (PC) analyses biplots of the soil metric data for sites along a rainfall gradient in dry (**a**) and rainy (**b**) seasons. Red circles 711-mm of mean annual rainfall sites; green triangles 917 mm of mean annual rainfall sites; blue squares 1232 mm of mean annual rainfall sites; lilac symbols 1611 mm of mean annual rainfall sites.


**Table 3.** Eigenvalues, cumulative percent variation, and eigenvectors of the first three principal components (PCs) for the soil metrics in the dry season.

**Table 4.** Eigenvalues, cumulative percent variation, and eigenvectors of the first three principal components (PCs) for the soil metrics in the rainy season.


The correlation matrix (Pearson's correlation test) across the rainfall gradient show a relevant pattern among soil C, N and P metrics impacted by the rainfall regime (Tables S1–S4). Consistently, two groups of interacting soil metrics appeared. The first group was composed by the sites at the driest and the wettest ends of the gradient with more significant associations between soil metrics (44 in the case of the site that receive 711 mm of rainfall, and 48 in the case of the site that receive 1661 mm of rainfall). The second group was composed by sites at the intermediate range of the rainfall gradient, that show a lower number of significant associations among soil metrics (20 and 35, for sites at 917 mm and 1232 mm of MAR, respectively).

#### **4. Discussion**

#### *4.1. E*ff*ects of Forest Conversion on Soil Carbon, Nitrogen and Phosphorus Dynamics*

The present study provides a quantitative overview of C, N and P concentrations, and metrics of their fluxes in the upper mineral soil layer under natural forests and converted livestock lands across the southeast of Mexico. First, we found a loss of organic C, total N and available P with land-use change. Although this organic C trend is consistent with the meta-analysis of Guo and Gifford [57], who reported that the accumulation of organic C in the mineral soil is altered with the conversion of forest cover to pastures, the significant variation in C values across sites following the conversion from natural forests to livestock production systems was probably due to differences in soil type and vegetation, and effects of land-use disturbance on soil parameters. Nonetheless, our study still suggests that land-use change alters the C-holding capacity of soil as regards short C retention capacity, and that intensive management practices (e.g., clear-cutting and slash burning for site preparation and pruning) and rainfall conditions could favor increased soil C losses [58]. On the other hand, the loss observed in the total N is unexpected, considering the inputs of N with the animal excreta in both livestock production systems. Moreover, although biological N fixation is the primary source of N input in tropical lands [59], the observed loss in soil total N at study sites after the conversion to intensive silvopastoral system indicates that land-use change did alter significantly the balance between N input and loss, with consequences for N retention in this agricultural ecosystem.

The significant differences in C and N observed at study sites due to forest conversion could be related to differences in both quantity and quality of C and N inputs through litter and different management practices between native forests and livestock production systems. Plant species differ in their C sequestration potential, and land-use change by changing woody tree species covers with recalcitrant C compounds [60] to non-woody pasture species or leguminous trees, would be expected to alter the sequestration potential of C. Besides the expected large differences in litter inputs to the soil between natural forests and N-rich litter in silvopastoral systems, forest floor mass and nutrients under the natural forests species had a large quantity of organic materials poorly decomposed that would be incorporated into the mineral soils, as has been observed in these forest ecosystems [61], while a high rate of litter decomposition is expected in silvopastoral systems due to high N concentration and low content of recalcitrant compounds in litter from leguminous trees [62].

Soil basal respiration has been used as a powerful index to evaluate the alterations of soil C cycle [63]. In the present study, land-use change increased the soil basal respiration suggesting a higher ratio of labile soil organic C (i.e., soil organic C pool with "fast" decomposition [64,65]) to the total soil organic C under pastures and *Leucaena* plantations than under natural forests, probably reflecting the expected differences across land uses in the quantity and/or quality of the C inputs. It is known that it is the ratio of labile soil organic C to total soil organic C, rather than the total soil organic C content itself, that influences soil quality [66], due to its key role for metabolically active microbes [67] and nutrient supply to growing plants [68]. These expected differences in the contribution of labile soil organic C to soil total organic C among land uses could be related to the chemical recalcitrance of root tissue in the studied native forests [60]. On the other hand, management involving no-till in livestock production systems allows the accumulation of labile forms of organic C in soil surface layers [69,70], while more stable organic C accumulates under native forest soils [71]. In addition, inputs of dung-derived C and N often induce microbial priming effects, promoting soil organic matter decomposition [72]. Aside from these possible drivers of changes in soil organic C mineralization, the observed differences in CO2 fluxes between natural forests and livestock systems could persist for an extended period considering that organic C stabilization in soil is relatively long-lasting (between 20 and 60 years; [73]).

Additionally, we observed a generalized loss in soil available P values after land-use change, despite the fact that dung is rich in P [74]. Moreover, the activity of the acid phosphatase under silvopastoral systems increased relative to its activity under natural forests, with largest changes in the

wettest end of the rainfall gradient. We cannot determine whether such increase in P-acquiring enzymes under intensive silvopastoral systems resulted from an increase in P-demand by *Leucaena* plantations (thus exacerbating the possible direct nutrient limitation to soil microbes). Generally, P supply limits ecosystem function in tropical zones [12] and in the study region [74]. The land-cover change of native forests to extensive pasture systems maybe does not change this situation. However, the intensive silvopastoral systems maybe makes P-limitation worse, as indicated also by the low availability of P in the soils. Although this hypothesis of pervasive P limitation in silvopastoral systems could be verified with a study of primary production at the ecosystem level, because microbial biomass C:N:P ratios are relatively invariant across ecosystems [75], differences in microorganism efforts to acquire P should be expected between ecosystems that differ in nutrient supply in order to maintain microorganism homeostasis [76], (i.e., soils poor in P have high levels of activity of P-acquiring enzymes, as was observed in the intensive silvopastoral systems).

#### *4.2. E*ff*ects of Rainfall Regime on Soil Carbon, Nitrogen and Phosphorus Dynamics*

Climatic variables have strong impacts on soil organic C, N and P contents and fluxes in tropical and extratropical soils [5,77], and previous studies suggested that rainfall is the primary factor controlling soil C, N and P dynamics in water-limited tropical ecosystems [15,78]. Our study indicates that organic C, total N and available P, and microbial biomass C and N concentrations increased with decreased rainfall in both natural forests and livestock production systems. This result is in line with those obtained by Burke et al. [79] who reported that rainfall clearly has a direct role regionally and globally in the amount of soil C and nutrients stored. These patterns indicate that C and nutrient accumulation reflect a strong decrease in decomposition rate with the decrease in MAR reflecting also enrichment in organic matter recalcitrance with decrease in rainfall amount [60,80]. Moreover, the soil basal respiration rates decreased with the decrease in rainfall amount, suggesting a moisture limitation for soil C mineralization. In addition, the net N transformations differed considerably among sites with changes in rainfall amount. However, in contrast with the pattern observed for the C mineralization across sites, the observed highest N transformations in soils from the driest end of the gradient, with the largest values in samples taken in the dry season, suggested that there is a large potential of N losses when the rainy season starts. Thus, the expected reduction in rainfall in the southeast of Mexico [81] could open the N cycle irrespective of land use. Thus, our study also suggests that climate change could affect moisture-sensitive biogeochemical processes, and a N limitation could occur in native forests and livestock production systems if drought intensity increases.

Overall, our study demonstrated that soils (both under native forests and livestock systems) at the driest end of the rainfall gradient had C, N and P accumulation during the dry season, probably due to the decrease in decomposition [62] and soil leaching during the rainless period. These large pools of resources in the dry season could be a large input of C and nutrients to microbial biomass and vegetation demands when the rainy season starts. Correlation analysis allowed us to identify variations in the interaction among soil C and nutrient dynamics across sites affected by both the amount and seasonality of rainfalls. In line with Finzi et al. [82], and Delgado-Baquerizo et al. [83] who propose that drought increase would uncouple the C, N and P cycles, our study indicates that the strength of relationships among soil metrics for C, N and P cycles decreased with the decreases in MAR from the 1661 mm rainfall site to the 917 mm rainfall site, indicating that the cycles of C, N and P in the soil are less coupled with drought. A recent study in Texas found that greater soil water content favored soil bacterial communities [84]. Contrary to this trend, the association between soil variables increase below 917 mm of MAR. This could be due to more intense and prolonged seasonal drought that limits soil leaching and favors the microbial control of soil C and nutrient dynamics [85–87].

Regarding the sustainability of the livestock sector in the southeast of Mexico, our study on the consequences of the land-use change and climatic variation in soil nutrient dynamics allows us to conclude that the regulatory mechanisms of soil fertility in these two grazing livestock systems will vary, depending on the details of the site's nutrient limitations and the expected drought intensification. Aside from these short- and long-term scenarios, the large extension of degraded land in the region [88] creates the opportunity for scaling up tropical forest restoration plans [89] or even the use of intensive leguminous tree plantations, which could be a positive contribution to reduce the massive impact of land use on tropical ecosystems, increase the landscape ecosystem services [8], and limit the contribution of the livestock sector to national N emission [90].

#### **5. Conclusions**

The analysis derived from our soil biogeochemical measurements provides an estimate of the response of tropical forest ecosystem and the sensitivity of land-use-change effects to increase in drought in the tropical region of Mexico. Our study indicates that expected climate change could impact moisture-sensitive biogeochemical processes, altering future carbon, nitrogen and phosphorus balances in soils under natural forests, and also in agricultural ecosystems, with potential negative consequences for the sustainability of livestock production systems in these tropical regions. Achieving the agricultural sustainability in these regions depends on limiting drought effects on soil carbon and nutrient losses with a better understanding of their biogeochemical dynamics as well as of the interactions among their cycles.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2071-1050/12/20/8656/s1, Table S1: Pearson correlation coefficients for soil metrics in sites that receive 711 mm of total annual rainfall, Table S2: Pearson correlation coefficients for soil metrics in sites that receive 917 mm of total annual rainfall, Table S3: Pearson correlation coefficients for soil metrics in sites that receive 1232 mm of total annual rainfall, Table S4: Pearson correlation coefficients for soil metrics in sites that receive 1661 mm of total annual rainfall

**Author Contributions:** Conceptualization—J.C., F.G.; formal analysis and data management—D.F., P.O.-F., T.F.A., A.R.-V., B.C.-V., J.D.E., J.C.; writing—original draft preparation—T.F.A., J.C.; writing—review and editing—J.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was made possible through support from the Mexican Consejo Nacional de Ciencia y Tecnología to D.F. and P.O.-F.; and by the Universidad Nacional Autónoma de México (PAPIIT grant 2015 RV200715).

**Acknowledgments:** We would like to thank Enrique Solís and Ofelia Beltrán-Paz for their valuable contribution on the soil analyses. We thank the landowners and managers for allowing access to their land and logistical help in the field.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Water Erosion Risk Assessment in the Kenya Great Rift Valley Region**

**George Watene 1,2,3,4, Lijun Yu 1,\*, Yueping Nie 1, Jianfeng Zhu 1,2 , Thomas Ngigi 3,4, Jean de Dieu Nambajimana 2,5 and Benson Kenduiywo <sup>3</sup>**


**Abstract:** The Kenya Great Rift Valley (KGRV) region unique landscape comprises of mountainous terrain, large valley-floor lakes, and agricultural lands bordered by extensive Arid and Semi-Arid Lands (ASALs). The East Africa (EA) region has received high amounts of rainfall in the recent past as evidenced by the rising lake levels in the GRV lakes. In Kenya, few studies have quantified soil loss at national scales and erosion rates information on these GRV lakes' regional basins within the ASALs is lacking. This study used the Revised Universal Soil Loss Equation (RUSLE) model to estimate soil erosion rates between 1990 and 2015 in the Great Rift Valley region of Kenya which is approximately 84.5% ASAL. The mean erosion rates for both periods was estimated to be tolerable (6.26 t ha−<sup>1</sup> yr−<sup>1</sup> and 7.14 t ha−<sup>1</sup> yr−<sup>1</sup> in 1990 and 2015 respectively) resulting in total soil loss of 116 Mt yr−<sup>1</sup> and 132 Mt yr−<sup>1</sup> in 1990 and 2015 respectively. Approximately 83% and 81% of the erosive lands in KGRV fell under the low risk category (<10 t ha−<sup>1</sup> yr<sup>−</sup>1) in 1990 and 2015 respectively while about 10% were classified under the top three conservation priority levels in 2015. Lake Nakuru basin had the highest erosion rate net change (4.19 t ha−<sup>1</sup> yr−1) among the GRV lake basins with Lake Bogoria-Baringo recording annual soil loss rates >10 t ha−<sup>1</sup> yr−<sup>1</sup> in both years. The mountainous central parts of the KGRV with Andosol/Nitisols soils and high rainfall experienced a large change of land uses to croplands thus had highest soil loss net change (4.34 t ha−<sup>1</sup> yr−1). In both years, forests recorded the lowest annual soil loss rates (<3.0 t ha−<sup>1</sup> yr<sup>−</sup>1) while most of the ASAL districts presented erosion rates (<8 t ha−<sup>1</sup> yr−1). Only 34% of all the protected areas were found to have erosion rates <10 t ha−<sup>1</sup> yr−<sup>1</sup> highlighting the need for effective anti-erosive measures.

**Keywords:** soil erosion; Great Rift Valley Lakes; ASAL; Kenya; desertification

#### **1. Introduction**

Soil erosion poses a serious threat to global agricultural production [1] with worldwide mean soil erosion rates and total annual soil loss estimated to be between 12 to 15 t ha−<sup>1</sup> yr−<sup>1</sup> and 2.5 to 4 billion tons [2], respectively. In East Africa (EA), particularly for countries within the east side of the Sudano-Sahelian region, rapid economic expansions resulting to unsustainable use of natural resources coupled with recent climatic changes have exacerbated on-site and off-site effects of soil erosion including flooding, environmental degradation and loss of agricultural land productivity [3–5]. Loss of productive soil by erosion in turn negatively impacts food security [1] as more than 50% of the population in most sub-Saharan countries depends on agriculture for their livelihood [6]. A multiple of

**Citation:** Watene, G.; Yu, L.; Nie, Y.; Zhu, J.; Ngigi, T.; Nambajimana, J.d.D.; Kenduiywo, B. Water Erosion Risk Assessment in the Kenya Great Rift Valley Region. *Sustainability* **2021**, *13*, 844. https://doi.org/10.3390/ su13020844

Received: 26 November 2020 Accepted: 11 January 2021 Published: 16 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

factors make tropical developing countries more vulnerable to the processes of soil erosion including high soil erodibilities, deforestation, desertification, agricultural intensification, poor soil conservation methods, and convergence of intense climatic regimes [7].

Desertification contributes nearly 80% of soil and land degradation in Kenya which has 88% of its land mass classified as arid and semi-arid lands (ASAL) [5]. Low vegetation cover, population increases combined with climatic changes has enhanced water erosion which further increases the risk of dryland degradation. These ASALs have been identified as most "vulnerable regions to climatic change" as well as the main cause of most socio-economic problems in the country [8]. With a total contribution of about 34.2% to the Gross Domestic Product (GDP-2019) [9], agriculture forms the mainstay of Kenya's economy. Mulinge et al. [10] estimated the total economic value of land degradation at 1.3 billion USD annually between 2001 and 2009 in Kenya which has over 12 million people living in degraded areas. Food crop productivity in Kenya's highly erodible soils [11] has in the recent past been hampered by soil erosion, increasing population and dynamic weather changes. Forest logging on mountain ranges and unchecked land-use activities along with intensive tropical precipitation increases soil erosion rates in Kenya's highland areas [12]. Water-induced soil erosion in the country's croplands has been estimated at 26 t ha−<sup>1</sup> yr−<sup>1</sup> [4] and has resulted to a permanent reduction of soil productivity in about 20% of its total area [10]. Reference [11] which is a recent Land and Degradation Assessment (LADA 2016) report in Kenya shows only about 2.2% of the total surface area has minimal risk of degradation while 61.4% suffers from severe degradation through soil erosion. In addition, the total forestland in Kenya decreased by 1% while croplands and bare lands (mostly in ASAL areas) increased by 7.3% and 2.6% respectively between 1990 and 2010 [11]. Land degradation thus threatens the source of the livelihoods of millions of Kenyans who predominantly depend on small scale farming [13]. In the central highlands of Kenya as with other most rural catchments, the total annual soil loss predictions vary from 134 t ha−<sup>1</sup> yr−<sup>1</sup> to 549 t ha−<sup>1</sup> yr−<sup>1</sup> which surpasses the estimated soil tolerance for tropical highland areas of 2.2–10 t ha−<sup>1</sup> yr−<sup>1</sup> [14,15]. The Great Rift Valley (GRV) region of Kenya which has 84.5% of its coverage classified as ASAL [16], in particular is prone to erosion due to steep topography, severe droughts, continued population pressure within its highland areas and dynamic land use and land cover changes over the last decade [17,18]. These changes include; overgrazing due to inadequate animal husbandry by pastoralists that leads to soil desertification [19], increased deforestation in the highland forest areas, e.g., Mau Forest and Cherangani Hills [17], massive shift from shrub lands to grasslands, human migration from lowlands areas (ASAL) to highland areas, and reduction in cropland areas. Land misuse coupled with climatic changes have made rills, gullies and sheet erosion are prevalent occurrence in the region, e.g., in the Baringo-Kerio valley [20]. The region is important as it forms the largest part of Kenya's grain basket zone that produces maize, wheat, beans, and tea [21]. The area has recorded high amounts of rainfall during the East African Short Rains (EASR) (October–December) periods in the recent past due to the Indian Nino resulting in rising lake levels in the GRV lakes [22]. Flooding of these lakes has also been attributed to a 50-year cyclic climatic event [23]. This has potentially heightened the risks of water erosion, pollution, siltation, flooding, and landslides [7,12].

A series of programs have been instigated in Kenya to address land degradation problems [10,21]. For instance, the forest policy has boosted reforestation measures which enabled the country to attain 6.99% national forest cover in 2014 [10]. Establishment of national parks and protected areas has reduced overgrazing and other adverse humanrelated activities in large rangelands although this significantly increases the pressure on the carrying capacity of the abutting lands. Some districts have adopted some Soil and Water Conservation (SWC) measures, e.g., conservation tillage practices and terracing, commonly referred to as *Fanya Juu* terraces, have shown considerable results [24]. With the current population density, the magnitude of soil erosion rates have continued to rise due to extreme weather changes, over-cultivation, desertification, and relative scarcity of productive farming land resulting to unsustainable sub division of land [3]. Since about

75 percent of the soils in Kenya are environmentally fragile and soil erosion is a longterm process involving complex combination of physical and hydrological factors [10,25], there is a need for methodical monitoring in order to establish critical areas and plan or devise for targeted soil conservation measures. Additionally, agriculture in Kenya strongly depends on irrigation water [21], and therefore it is important to monitor soil erosion spatial processes so as arm policy makers with efficient solutions to reduce soil transported into reservoirs.

Various studies have applied different models to compute soil loss by water erosion and sediment yields [26]. Some of the main hydrological models include the Universal Soil Loss Equation (USLE) [27], and its revised version (Revised Universal Soil Loss Equation (RUSLE)), Agricultural Non-Point Source Model (AGNPS), Morgan–Morgan–Finney (MMF), Soil and Water Assessment Tool (SWAT), Water Erosion Prediction Project (WEPP), Erosion Productivity Impact Calculator (EPIC), European Soil Erosion Model (EUROSEM), and The Limberg soil erosion mode (LISEM) [26]. These are broadly classified into physical and experiential models [28]. The applicability of a particular model generally depends on watershed spatial scale or characteristics, data accessibility and efficiency. Despite their complexity and high data requirements, physical models have "inbuilt process-based sub models" [26] that represent erosion processes more realistically. At regional scales and large catchments with limited data, empirical models such as USLE and RUSLE are most commonly applied [29] to estimate potential water erosion rates. The RUSLE is an updated version of the USLE model that has been extensively applied in many areas with different terrain characteristics and climatic zones to estimate long-term potential annual soil erosion rate, mainly because of its easy integration with geospatial technologies and low data requirements. Recent advancements in Geographic Information Systems (GIS) have enhanced RUSLE to allow for erosion monitoring at varied spatial and temporal scales [30].

Many of the past soil erosion studies in Kenya focus on the catchment scale or are at the local level [12,24,25,31]. Defersha et al. [32] applied WEPP and EROSION 3D physical models to estimate erosion rates and sediment yields at the Mara River Basin and revealed that the mean annual erosion rates for cultivated lands (120 t ha−<sup>1</sup> yr−1) was higher than that of bush land (7 t ha−<sup>1</sup> yr−1) or grasslands (3 t ha−<sup>1</sup> yr−1). Mati et al. [33] assessed the applicability of EUROSEM in two small catchments and found it inadequate for dry rangeland areas. Baker et al. [34] used the SWAT model in River Njoro watershed on the floor of Kenya's Rift Valley and showed that surface runoff increased proportionately with changes in land use. Similarly, Hunink et al. [35] indicated that coffee and maize growing areas presented mean erosion rates of 50 t ha−<sup>1</sup> yr−<sup>1</sup> and 10 t ha−<sup>1</sup> yr−<sup>1</sup> in the Upper Tana basin, respectively. In the USLE study by Mati et al. [24], 36% of the Upper Ewaso Ng'iro basin was predicted to suffer from mean erosion rates above the tolerable rate of 10 t ha−<sup>1</sup> yr−<sup>1</sup> mostly in the overgrazed rangelands. However, despite the presence of soil erosion in the physiographical regions of Kenya, few studies have applied the RUSLE model for spatial temporal evaluations particularly at the regional or national level [25,28,31,36]. The present study targets the GRV region of Kenya which covers about 33% of the country's total surface area with an aim to quantify (i) estimate the magnitude of potential soil loss rates in 1990 and 2015; (ii) assess the spatial changes among soil erosion risk classes between the two periods; (iii) identify priority areas for SWC; and (iv) quantify annual soil loss rates in Kenya Great Lakes ASAL basins, topography and protected areas.

#### **2. Materials and Methods**

#### *2.1. Study Area*

The Kenya Great Rift Valley (KGRV) region is located in the tropical zone of East African Rift System (EARS) and geographically lies between latitudes 4◦12- N and 3◦15- S and longitudes 34◦00- E and 38◦05- E (Figure 1). The region shares its northern border with Sudan–Ethiopia border, southern border with Tanzania and about half of its western border with Uganda. It has a total area of 194,291.73 Km2, corresponding to approximately 33% of the country's total area The area is characterized by undulating volcanic and tectonic terrain with altitudes ranging from 360 to 4170 m.a.s.l with a mean altitude of approximately 1200 m.a.s.l. The Eastern Rift Valley traverses north south across the region approximately 720 km and 110 km wide. The Kenya Lake System UNESCO World Heritage Site is located in this region [37] and Lake Turkana which is located in the north is both the world's largest permanent desert lake and largest alkaline lake. The area encompasses Kenya's four main water towers: the Mau Forest Complex which supports an important ecosystem—including an equivalent market value of 229 million USD for the tea and tourism economic sectors only [38]—Cherangani Hills and sections of Mt. Elgon and Aberdare Ranges. These vital ecosystems face constant threats from both anthropogenic forces (encroachment and deforestation) and natural hazards that have resulted to drying up of some rivers and streams within the region. Figure 2 shows Landsat time-series images indicating rising water levels in Lake Baringo due increased rainfall in the East Africa (EA) region (EASR) [22] and large-scale deforestation within the Mau forest Complex highlands [17].

**Figure 1.** Map of the study area: (**a**) administrative districts and Agro-Ecological Zones (AEZ) [16,21] within the Great Rift Valley region of Kenya and (**b**) the location of Great Rift Valley region in Kenya (**c**) location of Kenya in the Africa continent.

**Figure 2.** Landsat time-series images on selected areas in the Kenya GRV region: (**a**,**b**) indicating rising water levels in Lake Baringo between 30th May, 2013 and 1st May, 2020 respectively; (**c**,**d**) indicating large-scale deforestation in the Mau Forest Complex between 1st February, 2002 and 16th February, 2016 respectively.

The area has a tropical climate with a mean annual precipitation of about 614 mm and two wet seasons (March–May (*Masika*) and September–December (*Vuli*)). The highland areas including Mau Forest (2000 to 2800 m above sea level) enjoy high intensity rainfall ranging between 1000 and 2000 mm and mean annual temperatures of 10 and 22 ◦C. The long-term (1970–2015) mean annual precipitation of Lodwar (8635000), Egerton University (KE0863), and Narok (9135001) meteorological stations representing Upper, Central, and Lower climatic zones of the KGRV, respectively, ranges from about 5 mm to 150 mm as shown in Figure 3. Daily rainfall data from the Kenya Meteorological Department (KMD) showed that Lodwar, Egerton University, and Narok stations recorded 2, 26, and 14 days with rainfall measurement greater than 10 mm, respectively, in the year 1990. Similarly and following the same order, the stations recorded 6, 28, and 19 days in the year 2015. The maximum daily rainfalls for Lodwar, Egerton University, and Narok stations in 1990 are 30.1, 72.2, and 46.5 mm, respectively, and 30.8, 41.4, and 36.2 mm, respectively, in 2015. The northern Lotikipi plains in Turkana district experience low amounts of rainfall ranging from 3 to 55 mm yearly with mean annual temperatures varying from 28 to 31 ◦C. The tropical highlands of the KGRV are mostly associated with Andosols and Nitisols soils that are developed from volcanic material. Cambisols are within areas with medium elevation while Lithosols, Solonchaks, and Regosols are prevalent in the ASAL regions. The dominant soil categories (Figure 4) include Lithosol (29.8%), Regosols (15.0%), Nitosols (10.4%), Cambisols (7%), and Ferrasols (6.8%) [39]. The area has five agro-ecological zones (AEZ) (Figure 1) include Arid North (with a mean annual rainfall of 506 mm), Semi-Arid North (759 mm), Semi-Arid South (762 mm), High Rainfall (1188 mm), and Turkana (258 mm) with a proportion of each zone contributing 24.8%, 9.5%, 20.5%, 15.5%, 29.7%, and 15.5%, respectively, of the total region area [16,21,40]. The region presents ideal conditions to conduct soil loss analysis for the country due to its unique environmental diversity (i.e., combination of ASAL and High Rainfall AEZ) that covers four of the five drainage basins

in Kenya [40]. Topography and landforms largely shapes the region's drainage pattern (Figure 4). Several rivers branch from the central Kenya highlands into the endorheic Great Rift Valley basin, rivers in the western areas flow westward into Lake Victoria while streams from the Sudan-Ethiopia border drain into Lake Turkana (Figure 4). The lithology of the region is mainly dominated by igneous rocks around the mountainous landforms with sedimentary and metamorphic rocks mainly occupying the northern and western parts respectively. The four dominant landform types include Plains (29.2%), Plateaus (17.8%), Hills and Mountain foot ridges (12.6%), and Mountains (10.4%). Based on the 2019 national census the area has an estimated population of 13.8 million [9] which derive its livelihood mainly from agriculture in the High Rainfall AEZ and animal husbandry in the ASALs.

**Figure 3.** Average monthly rainfall (mm) for the Kenya Great Valley Region.

#### *2.2. Data Collection*

Table 1 shows the key datasets used in this study obtained from different sources. All the data was re-projected to the World Geodetic System (WGS) 1984\_Universal Traverse Mercator (UTM) and resampled to match the data with coarse spatial resolution (250 m) using the SDMtoolbox in ArcGIS 10.5 software.

#### *2.3. Land Use and Land Cover (LULC) Maps*

For this study, 1990 and 2015 land use and land cover (LULC) maps for the GRV in Kenya (Figure 5) were acquired from the Department of Resource Surveys and Remote Sensing (DRSRS), Kenya in order to mask out non-erodible areas, estimate erosion rates for different LULC categories [41] and the impact of land use land cover changes (LULCC) on soil los rates. Erosion-prone areas covered a total area of 185,884.3 Km2 that included Dense Forest (4.9%), Open Forest (3.3%), shrub land (48.6%), grassland (21.6%), cropland (4.9%), and bare lands (16.7%), in 1990. Following a similar order, the proportion of land uses were 4.2%, 3.0%, 45%, 20.7%, 11.8%, and 15.3% in the year 2015.

**Figure 4.** Maps of the Kenya Great Rift Valley (KGRV) region: (**a**) Landform types (**b**) Dominant Soil types region in Kenya (**c**) Lithological map (**d**) Major river basins.

**Figure 5.** Land use/land cover distribution maps for the Kenya Great Rift Valley Region: (**a**) 1990 (**b**) 2015.


**Table 1.** The key datasets used for the Revised Universal Soil Loss Equation (RUSLE) model in the KGRV.

DEM—Digital Elevation Model; SRTM GL1—Shuttle Radar Topography Mission Global 1 arc second; NASA— National Aeronautics and Space Administration, US; AfSIS—Africa Soil Information Service; ISRIC-WSI— International Soil Reference and Information Centre, World Soil Information; MoE&F(K)—Ministry of Environment and Forestry (Kenya); CHIRPS—Climate Hazards Group InfraRed Precipitation with Station data; USGS-EROS—United States Geological Survey, Earth Resource Observation Center.

#### *2.4. RUSLE Model Application*

Due to its unique merits that include easy integration with geospatial technologies [30], applicability in areas with limited data and adaptability at different spatial scales the Revised Universal Soil Loss Equation (RUSLE) empirical model has broadly been applied to estimate soil erosion rates worldwide.

The RUSLE model was chosen in this study since it has been tested in different landscapes, its application expediency and low data requirements [45]. The RUSLE equation [45] (Equation (1)) incorporates five different environmental variables using geoinformatics techniques to estimate the characteristics of soil erosion (Figure 6). These factors are rainfall erositivity (*R*), soil erodibility (*K*), slope length and steepness (*LS*), cover management (*C*), and support practice (*P*) [27].

$$A\_{\perp} = R \ast K \ast LS \ast \mathbb{C} \ast P \tag{1}$$

where *A* = annual average soil loss (t ha−<sup>1</sup> yr−1); *R* = rainfall erosivity factor (MJ mm ha−<sup>1</sup> h−1y−1); *K* = soil erodibility factor (t ha h ha−<sup>1</sup> MJ−1mm−1); *LS* = slope length and slope steepness factor (dimensionless); *C* = cover management factor (dimensionless); *P* = support practice factor (dimensionless).

**Figure 6.** The methodological framework for estimating water-induced soil erosion rates using RUSLE model in the Great Rift Valley region of Kenya. LCLU: Land Use and Land Cover, DRSRS: Department of Resource Surveys and Remote Sensing, CHIRPS: Climate Hazards Group InfraRed Precipitation with Station data, SRTM: Shuttle Radar Topography Mission, AfSIS: Africa Soil Information Service.

#### *2.5. Determination of RUSLE Factors*

#### 2.5.1. Rainfall Erosivity (R) Factor

Due to the direct correlation between rainfall intensity and erosion, the *R* factor represents the main driving factor of soil erosion [30] and contributes almost 80% of total soil loss [46]. The classical Wischmeier and Smith (1978) calculation method for *R* factor requires the use of storm erosivity index (EI) values of at least 20 years to account for seasonal variabilities and rainfall intensities [14,27]. Globally, not many areas have such gauged data readily available especially in developing countries [28]. The Lo et al. [47] method (Equation (2)) was adapted to estimate the rainfall erosivity factor since it has been applied by several studies in the East Africa region with significant results [29,48].

$$R = \text{--} 38.46 + \text{3.48} \times P \tag{2}$$

where *R* = Rainfall Erosivity in MJ mm ha−<sup>1</sup> h−<sup>1</sup> yr−1, *P* is the mean annual precipitation in mm. The mean annual precipitation for the time periods 1981–1999 and 1999–2015 were computed from the monthly average precipitation downloaded from the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) [44] database and used to calculate rainfall erosivity factors for the periods 1990 and 2015, respectively [31]. CHIRPS data is readily available at high spatial-temporal resolutions and has been shown

to have significant results in determining long-term rainfall trends when compared to rain gauge station datasets for Kenya [49,50] as well as the entire East Africa region [7,51].

#### 2.5.2. Soil Erodibility (K) Factor

The *K* factor indicates the ability or resistance of soil particles to disintegrate and be transported by surface water runoff. This is dependent on the inherent soil properties including soil texture, organic matter, soil structure, and permeability [52]. To determine the *K* factor, the EPIC (erosion-productivity impact calculator) [53] model was applied to the sand, organic, silt and sand soil fractions of the area as compiled by the Africa Soil Information Service (AfSIS) [43].

$$K = F\_{csand} \times F\_{si-cl} \times F\_{orgc} \times F\_{hisand} \times 0.1317,\tag{3}$$

where

$$F\_{\rm sand} = \left[ 0.2 + 0.3 \exp\left( -0.0256 \text{S} \text{A} \text{N} \left( 1 - \frac{S \text{II}}{100} \right) \right) \right] \tag{4}$$

$$F\_{si-cl} = \left[\frac{SIL}{\mathbb{C}LA + SIL}\right]^{0.3},\tag{5}$$

$$F\_{\text{orgc}} = \left[1.0 - \frac{0.0256 \text{C}}{\text{C} + \exp(3.72 - 2.95 \text{C})}\right] \text{.} \tag{6}$$

$$F\_{\text{hisand}} = \left[1.0 - \frac{0.70 \text{ SN1}}{\text{SN1} + \exp(-5.51 + 22.9 \text{ SN1})}\right] \tag{7}$$

where *SAN, SIL,* and *CLA* are percent sand, silt and clay content, respectively; *C* is the organic carbon content; and *SN1* is sand content subtracted from 1 and divided by 100. *Fcsand* (Equation (4)) = gives a low soil erodibility factor for soil with coarse sand and a high value for soil with little sand content. *Fsi-cl* (Equation (5)) = gives a low soil erodibility factor with high clay to silt ration *Forgc* (Equation (6)) = is the factor that reduces soil erodibility for soil with high organic content. *Fhisand* (Equation (7)) = is the factor that reduces soil erodibility for soil with extremely high sand content.

#### 2.5.3. Slope Length and Slope Steepness (LS) Factor

The *LS* factor represents the impact of topography on soil erosion [54]. It expresses the effects of local landscape on soil loss and is taken as the product of two terrain attributes: slope length (*L*) and slope steepness factor (*S*). Increasing the slope length and slope steepness values leads to higher overland flow speed and accelerates erosion rates [48]. The (*LS*) can be defined as the ratio of soil loss on a given slope length and steepness to soil loss from a seedbed with 22.13 m slope length and a steepness of 9% where all other conditions are held constant [54]. SRTM GL1 version 3 (30 m resolution) dataset provided by the United States Geological Survey (U.S.G.S.) [42] for the region was acquired to derive terrain attributes using the Raster Calculator tool from the Spatial Analyst extension of ArcMap 10.5 (Environment Systems Research Institute (Esri) Inc., Redlands, CA, USA). The *L* factor was calculated following algorithm (Equation (8)) proposed by Desmet and Govers (1996) while the S factor was estimated using the McCool et al. (1987) method (Equation (11)) [55,56]

$$L\_{i,j} = \frac{\left(A\_{i,j-in} + D^2\right)^{m+1} - A\_{i,j-in}^{m+1}}{D^{m+2} \cdot x\_{i,j}^m \cdot \left(22.13\right)^m},\tag{8}$$

$$m = \frac{\beta}{1 + \beta} \tag{9}$$

$$\beta = \frac{\sin \theta / 0.0896}{3 \left( \sin \theta \right)^{0.8} + 0.56} \tag{10}$$

$$S\_{i,j} = \begin{cases} 10.8 \sin \theta\_{i,j} + 0.03, \tan \theta\_{i,j} < 9\% \\ 16.8 \sin \theta\_{i,j} - 0.50, \tan \theta\_{i,j} \ge 9\% \end{cases} \tag{11}$$

where *Li.j* = slope length factor for the grid cell with coordinates *(i.j)*; *D* = the grid cell size (*m*); *Xi.j* =; *ai.j* = aspect direction for the grid cell with coordinates (*i.j*); *Ai.j-in* = Flow accumulation or contributing area at the inlet of a grid cell with coordinates (*i.j*) (*m2*), *β* = the ratio of inter-rill erosion, *θ* = the slope in degrees [56].

#### 2.5.4. Cover Management Factor (C) Factor

The *C* factor corresponds to impact of vegetation canopy and land management practices on soil loss [27]. Reference [27] defined the *C* factor as the proportion of soil loss from land cropped under specific conditions to the corresponding soil loss under cleantilled, continuous fallow land. The cover management factor ranges from 0 for non-erosive areas with thick vegetation cover to 1 which indicates very high susceptibility to erosion due to intensive tillage or exposed smooth surfaces. Nyssen et al. [57] emphasized on the importance of *C* factor in soil erosion assessments thus misrepresentations of *C* factor coefficients for different land covers can result in high over or under estimations of erosion rates. For this study, *C* value coefficients were adopted from different literatures mainly focusing on the East Africa region (Table 2) and assigned to the corresponding thematic LULC raster maps.

**Table 2.** Adopted *C* values for different land use patterns (after past literatures in East Africa (EA)).


#### 2.5.5. Support Practice (P) Factor

The *P* factor represents land management control practices aimed at decreasing the rate of surface water runoff which in turn reduces soil erosion [27,56]. Conventional conservation measures include contouring, strip-cropping and terracing. Determining *P* factor values at large regional scales is nontrivial due to scarcity of data regarding conservation practices as well as complexities presented by different land uses [28]. A maximum value of "1" indicating poorest conservation practices was set to the P-factor [27] for the entire region since precautionary measures are often overlooked in regional soil erosion investigations [4]. In addition, Reference [64] revealed that the impact of soil and water conservation measures rapidly diminish in East African semi-arid areas. This study also separately estimated mean erosion rates in agricultural areas within the central and southern parts of the Kenya GRV region (Figure 1) in the year 2015 taking into account the three traditional soil erosion control practices proposed by Shin (1999) [65] (Table 3) to better understand their effect on croplands.

**Table 3.** *P* factor estimates for the common erosion control practices focusing on slope (%) [65].


#### **3. Results**

#### *3.1. Estimated Soil Erosion Rates in the Great Rift Valley Region of Kenya*

Figures 7 and 8 show the RUSLE results for the two periods while their statistical details are provided in Table 4. Rainfall Erosivity Factor (*R*) value varied between 359 and 8241 MJ mm ha−<sup>1</sup> h−<sup>1</sup> yr−<sup>1</sup> (Figure 7a) with a mean of 2626.7 MJ mm ha−<sup>1</sup> h−<sup>1</sup> yr−<sup>1</sup> in the year 1990. For the year 2015, the value varied from 340 and 7974.9 MJ mm ha−<sup>1</sup> h−<sup>1</sup> yr−<sup>1</sup> with an average of 2162 MJ mm ha−<sup>1</sup> h−<sup>1</sup> yr<sup>−</sup>1. In both years the highland areas recorded rainfall erosivity values >5000 MJ mm ha−<sup>1</sup> h−<sup>1</sup> yr−<sup>1</sup> as with the work of Reference [7]. The central areas of the Kenya Great Rift Valley region is largely dominated by high *R* value while at the upper and lower parts (ASAL zones), *R* values are in the low range. These *R* values are consistent with the spatial distribution of the average annual rainfall across the region. Soil erodibility Factor (*K*) values ranged from 0.014 to 0.026 t ha h ha−<sup>1</sup> MJ−<sup>1</sup> mm−<sup>1</sup> (Figure 8a). The topographic factor (*LS*) values were classified into five categories (Figure 8b) while the cover management Factor (*C*) ranged between 0 and 0.4 (Figure 7d).

The mean erosion rate for the year 1990 was estimated at 6.26 t ha−<sup>1</sup> yr−<sup>1</sup> with a standard deviation of 50.71 while the year 2015 presented a rate of 7.14 with a standard deviation of 40.38. In both years, the estimated mean rate of annual soil loss fell within the normal soil loss tolerances (from 5 to 11 t ha−<sup>1</sup> yr<sup>−</sup>1) [14,27,46]. The amount of total annual soil loss in the KGRV region was 116 Mt yr−<sup>1</sup> in 1990 and 132 Mt yr−<sup>1</sup> in 2015. To show the spatial distribution of water erosion and their areal extents in 1990 and 2015, the study area was classified into six erosion risk categories ranging from "very low" to "extremely high". In both years, the very low and low erosion classes when combined constitute greater sections of the total study area. In 1990, the two classes totaled to 154,822.1 Km2 (83.3% of the total study area) while in 2015, the total was 150,695.2 Km2 (81.1% of the total study area). The areal extent covered by medium, high medium, high, very high, and extremely high increased from 16,249.6 Km2, 10,783.2 Km2, 3 564.3 Km2, 257.9 Km2, and 207.0 Km2 in 1990 to 16,906.1 Km2, 11,838.3 Km2, 5430.8 Km2, 561.4 Km2, and 452.4 Km2, respectively, in 2015. However, the low erosion class reduced from 99,265.1 Km2 to 93,733.1 Km<sup>2</sup> over the study period.

**Table 4.** The erosion risk classes and their corresponding erosion rate net changes between the 1990 and 2015 periods.


**Figure 7.** The Revised Universal Soil Loss Equation (RUSLE) factor maps of the KGRV: (**a**) rainfall erosivity factor, 1990 (supplementary materials); (**b**) cover management factor, 1990 (supplementary materials); (**c**) rainfall erosivity factor, 2015 (supplementary materials); and (**d**) cover management factor, 2015 (supplementary materials).

**Figure 8.** The Revised Universal Soil Loss Equation (RUSLE) factor and results maps of the KGRV: (**a**) soil erodibility factor (supplementary materials); (**b**) slope length and slope steepness factor (supplementary materials); (**c**) spatial distribution of man annual soil loss in 1990 (supplementary m aterials); and (**d**) spatial distribution of man annual soil loss in 1990 (supplementary materials).

#### *3.2. Land Use/Land Cover Changes (LULCC) and Soil Erosion in the Great Rift Valley Region of Kenya*

Table 5 shows the comparison of soil loss estimates between changes in LULC types within the investigated area. The results suggest a significant incline of mean rate of soil loss in cropland (from 15.8 t ha−<sup>1</sup> yr−<sup>1</sup> in 1990 to 20.6 t ha−<sup>1</sup> yr−<sup>1</sup> in 2015) and dismal reduction across all the other LULC types. LULCC occurred in the area of about 71,044.6 Km2 (38.2%), while 114,839.7 Km<sup>2</sup> (61.8%) remained unchanged over the study period. Notable LULC types experiencing conversions (Table A1) include shrubland to grassland 15,077.3 Km2, grassland to shrubland 12,289.9 Km2, grassland to cropland 7530.7 Km2, bareland to grassland 4686.5 Km2, and dense forest to cropland 1643.6 Km2. Among LULCC that contributed to increased soil losses are dense forest to bareland (59.0 t ha−<sup>1</sup> yr−1), dense forest to cropland (31.0 t ha−<sup>1</sup> yr−1), open forest to cropland (31.0 t ha−<sup>1</sup> yr−1), shrubland to cropland (10.7 t ha−<sup>1</sup> yr−1), grassland to cropland (10.2 t ha−<sup>1</sup> yr−1), and open forest to grassland (5.7 t ha−<sup>1</sup> yr<sup>−</sup>1). Conversely, LULCC that contributed to reduced soil losses include cropland to dense forest (−26.4 t ha−<sup>1</sup> yr−1), shrubland to open forest (−10.1 t ha−<sup>1</sup> yr<sup>−</sup>1), and shrubland to dense forest (−19.5 t ha−<sup>1</sup> yr<sup>−</sup>1) (Table 5).

**Table 5.** Estimated mean erosion rates per LULCC (Land use/land cover change) category (1990– 2015) for Dense Forest, Open Forest, Shrubland, Grassland, Cropland and Bareland.


Spatial analysis of erosion risk conducted at the district level revealed that most of the districts within the mountainous landform part of the KGRV recorded soil loss rates >10 t ha−<sup>1</sup> yr−<sup>1</sup> (Table A2). Kericho district was consistent in presenting mean rates greater than 20 t ha−<sup>1</sup> yr−<sup>1</sup> in both study periods while most ASAL (e.g., Turkana, Laikipia and Kajiado) districts had rates <5 t ha−<sup>1</sup> yr−1. Keiyo and Marakwet districts had the highest soil loss increment of about 10 t ha−<sup>1</sup> yr−<sup>1</sup> over the study period. This is agreement with work of Reference [31] who reported an average increase of about 6 t ha−<sup>1</sup> yr−<sup>1</sup> for the western parts of the KGRV. This can be attributed to the high altitude and high mean rainfall which favored intensive farming on highly erosive soils in the region [31]. Table A3 shows the distribution of LULCC in relation to soil erosion at a district level. It represents the area coverage of the LULCC that occurred per district and the corresponding mean soil loss rates. The results suggest that though Keiyo district experienced the least conversions, it had highest erosion rates due to its high slope (19.8%) and mean rainfall (about 1200 mm). On the contrary, the driest districts characterized by semiarid desert plateaus recorded low erosion rates despite their high LULC conversion exchanges (e.g., Turkana and Samburu).

#### *3.3. Estimated Soil Erosion Rates in the Protected Areas within the Great Rift Valley Region of Kenya*

The United Nations Environment Program (UNEP) and the World Conservation Monitoring Center (WCMC) puts the total number of protected areas of Kenya at 411 with coverage of approximately 72,545 Km<sup>2</sup> [28,66] The number of the protected areas listed in the Great Rift Valley region of Kenya (Table A4) with a soil rate of <10 t ha−<sup>1</sup> yr−<sup>1</sup> went down from 57% in 1990 to 34% in 2015. Most of the protected areas that recorded an inclined of mean erosion rate are located in the areas occupied by characterized steep gradients topography and high rainfall intensity. Some of the endangered areas with mean erosion rates >35 t ha−<sup>1</sup> yr−<sup>1</sup> in year 2015 that need intervention planning include Kessop (51.86 t ha−<sup>1</sup> yr<sup>−</sup>1), Sogotio (40.7 t ha−<sup>1</sup> yr<sup>−</sup>1), Kaisungor (47.04 t ha−<sup>1</sup> yr<sup>−</sup>1), Chemurokoi (40.87 t ha−<sup>1</sup> yr−1), and Kimojoch (41.45 t ha−<sup>1</sup> yr−1). Internationally acclaimed areas like Masai Mara, Lake Nakuru, and Amboseli national parks had consistent low soil loss averages of 1.5 to 4 t ha−<sup>1</sup> yr−<sup>1</sup> in both years of study and dismal changes. Nevertheless, areas that experienced high cases of deforestation between the two periods presented sharp incline erosion rates, e.g., Eastern Mau (from 5.94 to 18.18 t ha−<sup>1</sup> yr<sup>−</sup>1), South Western Mau (from 4.56 to 13.45 t ha−<sup>1</sup> yr<sup>−</sup>1), Southern Mau (from 13.13 to 19.22 t ha−<sup>1</sup> yr<sup>−</sup>1), Kipkabus (from 17.77 to 34.15 t ha−<sup>1</sup> yr<sup>−</sup>1), and Timboroa (from 14.17 to 24.55 t ha−<sup>1</sup> yr<sup>−</sup>1).

#### *3.4. Classification Estimated Mean Erosion Rates by Severity and Conservation Priority*

To prioritize for conservation planning, the quantitative soil erosion loss map loss of the Great Rift Valley region of Kenya were classified into 6 erosion classes following the methodology by Koirala et al. [67] in order to identify conservation priority areas (Table 6). The erosion severity ordinal classes are namely: slight (0–5 t ha−<sup>1</sup> yr−1), moderate (5–10 t ha−<sup>1</sup> yr−1), high (10–20 t ha−<sup>1</sup> yr−1), very high (20–40 t ha−<sup>1</sup> yr−1), severe (40–80 t ha−<sup>1</sup> yr−1), and very severe (>80 t ha−<sup>1</sup> yr−1). Areas with very severe erosion levels have been categorized as first priority whereas slight erosion values allocated 6th conservation priority. The results show that areas under slight erosion decreased from 71.73% of the total erosive lands in 1990 to 69.46% in 2015. However, the extent of total erosive lands for moderate, high, very high, severe and very severe classes increased from 11.57%, 8.74%, 5.8%, 1.91%, and 0.25% in 1990 to 11.63%, 9.09%, 6.36%, 2.92%, and 0.54%, respectively, in 2015.

**Table 6.** The distribution of estimated soil erosion rates per different severity classes.


#### *3.5. Estimated Soil Erosion Rates by Slope and Elevation*

The mean annual soil loss for areas with high slopes (β = 17.6–26.8%) was 14.99 t ha−<sup>1</sup> yr−<sup>1</sup> resulting in a total loss of approximately 19.2 Mt yr−<sup>1</sup> in the year 1990 (Table 7). The erosive lands within this slope category had the largest erosion net change in the year 2015 which presented a rate of 18.53 t ha<sup>−</sup>1yr−1. Total soil loss for gentle slopes (β < 7%) increased slightly from 20.1 Mt yr−<sup>1</sup> to 21.1 Mt yr−1. The elevation raster map of the investigated area was reclassified into five different categories and the corresponding mean soil loss rates extracted using the ArcGIS Spatial Analyst tool set. The mean erosion rate for elevation of <500 m.a.s.l were 3.42 t ha−<sup>1</sup> yr−<sup>1</sup> in 1990 and 2.32 t ha−<sup>1</sup> yr−<sup>1</sup> in 2015 (Table 8). Areas with elevation greater than 2000 m.a.s.l characterized by high mean rainfall had the highest net change in soil loss (5.87 t ha−<sup>1</sup> yr−1) while regions with an elevation less than 1500 m.a.s.l having slightly reduced erosion rates.


**Table 7.** The distribution of estimated soil erosion rates by slope.

**Table 8.** The distribution of estimated soil erosion rates by elevation.


#### *3.6. Soil Erosion in the Major River Basins*

The Great Rift Valley region of Kenya covers four of the five major basins that drain in Kenya [40,68]: the entire Great Rift Valley Area basin (GRVA) and sections of Ewaso Ngiro, Lake Victoria and Athi River basins (Figure 4) that had mean annual soil rates of 6.16 t ha−<sup>1</sup> yr−1, 5.09 t ha−<sup>1</sup> yr−1, 9.42 t ha−<sup>1</sup> yr−1, and 3.0 t ha−<sup>1</sup> yr−<sup>1</sup> in 1990 and 6.79 t ha−<sup>1</sup> yr−1, 4.37 t ha−<sup>1</sup> yr−1, 13.7 t ha−<sup>1</sup> yr−1, and 3.6 t ha−<sup>1</sup> yr−<sup>1</sup> in 2015, respectively (Table 9). Of all the major sub basins within the KGRV region, only Lake Bogoria-Baringo had a mean annual soil loss rate >10 t ha−<sup>1</sup> yr−<sup>1</sup> in 1990 while Lake Victoria, Lake Nakuru, Lake Naivasha, and also Lake Bogoria-Baringo recorded soil loss rates >10 t ha−<sup>1</sup> yr−<sup>1</sup> in the year 2015.

**Table 9.** The estimated soil erosion rates and the corresponding net changes in the major river basins of the KGRV.


#### *3.7. Estimated Soil Erosion Rates by Major Landform and Soil Types within the KGRV*

Tables 10 and A4 present the statistical details of the soil loss rates within the major landform types and dominant soils within the Kenya Great Rift Valley region respectively. In both years of study, areas around escarpment had the highest mean erosion rates of >25 t ha−<sup>1</sup> yr−<sup>1</sup> with plains recorded lowest rates of about 2.5 t ha−<sup>1</sup> yr−1. Mountainous parts around the central KGRV had the highest incline in mean erosion rates (4.34 t ha−<sup>1</sup> yr−1) while depressions had decreased soil rates (−1.43 t ha−<sup>1</sup> yr−1) over the

study period. The results show that all of the dominant soil groups had mean soil loss rates <10 t ha−<sup>1</sup> yr−<sup>1</sup> in the year 1990 while Andosols and Nitisols (soils associated with volcanic material) recorded 11.53 t ha−<sup>1</sup> yr−<sup>1</sup> and 13.9 t ha−<sup>1</sup> yr−<sup>1</sup> in 2015, respectively. In addition, the two soil groups presented high soil loss rate net changes over the study period with Andosols (with high agricultural production) increasing by 5.52 t ha−<sup>1</sup> yr−1. Despite their resistance to water erosion due to good aggregate stability and high water permeability, Andosols can become susceptible to erosion particularly in intensively cultivated and deforested areas [69]. The heavy clayey Vertisols and the mostly water logged Gleysols had consistent low soil loss rates (about 3.0 t ha−<sup>1</sup> yr<sup>−</sup>1) in both years.


**Table 10.** The estimated soil erosion rates per landforms categories in the Great Rift Valley region of Kenya.

#### *3.8. Sensitivity Analysis of the RUSLE Model Factors used in the KGRV*

Each of the five RUSLE parameters has a different role or impact on the total magnitude of the mean erosion rate [52]. The descriptive statistics in the model shown in Table 11 revealed that rainfall erosivity parameter (*R*) and soil erodibility parameter (*K*) are the two strongest controlling parameters for soil erosion in the Kenya GRV region. Regions with severe erosion rates increased significantly after removing parameter *K*.



*3.9. Estimated Soil Erosion Rates in the Agricultural areas within the Central and Southern Rift Valley Region of Kenya (in 2015)*

The central and southern region that includes the High Rainfall AEZ where agriculture is mostly practiced within the study area [21] (Figure 8), was separately analyzed in 2015 to understand the effect of common conservation practices on cropland areas (i.e., contouring, strip-cropping, and terracing). This can assist in making risk-informed decisions to conserve croplands that substantially contributed to increased soil loss rates over the period of study. Under the baseline conditions with *P* factor values set to "one", the total cropland area (21,864.2 Km2) Figure 9a had a moderate mean soil erosion rate of 18.0 t ha−<sup>1</sup> yr−1; only 4.5% of the croplands had a sustainable mean soil loss <1 t ha−<sup>1</sup> yr−<sup>1</sup> while 28.6% of the croplands had severe soil loss rates >20 t ha−<sup>1</sup> yr−<sup>1</sup> mostly in central highland areas. This estimated mean annual soil loss rate is higher than the normal soil tolerances (from 5 to 11 t ha−<sup>1</sup> yr<sup>−</sup>1) [27] and the highland threshold for agro-ecological zones in tropical areas (from

0.2 to 11 t ha−<sup>1</sup> yr<sup>−</sup>1) [48]. The study predicted that, compared to the baseline scenario, terraces Figure 9b would decrease mean soil loss by 84.4% (from 18.0 to 2.8 t ha−<sup>1</sup> yr−1) while stripcropping Figure 9c and contouring Figure 9d would slow soil erosion rate by approximately 2.5 and 1.2 times (from 18.0 to 7.2 t ha−<sup>1</sup> yr<sup>−</sup>1) and (from 18.0 to 14.4 t ha−<sup>1</sup> yr<sup>−</sup>1), respectively.

**Figure 9.** The estimated mean soil erosion rates for croplands (21,864.2 Km2) within the central and southern regionsof the KGRV under different conservation practices in 2015 period: (**a**) baseline scenario; (**b**) terracing; (**c**) strip-cropping; and (**d**) contouring.

#### **4. Discussion**

#### *Overview of Estimated Soil Erosion Risk in the Great Rift Valley Region of Kenya*

The present study found that the mean erosion rate for the entire area was estimated at 6.26 t ha−<sup>1</sup> yr−<sup>1</sup> with a total soil loss of 116 Mtyr−<sup>1</sup> in 1990 (Figure 6) and 7.14 t ha−<sup>1</sup> yr−<sup>1</sup> with a total soil loss of 132 Mtyr−<sup>1</sup> in 2015. This estimated rate is within the range of erosion rate for Africa (10.8–146 t ha−<sup>1</sup> yr<sup>−</sup>1) [70] and slightly below the tolerable limits for mountainous environments (below 25 t ha−<sup>1</sup> yr<sup>−</sup>1) [67]. The mean erosion rate is also within range of the normal soil loss tolerances (ranging between 5 and 11 t ha−<sup>1</sup> yr<sup>−</sup>1) [27,48]. For both years, a greater proportional of the investigated region fell under the tolerable category (Table 4) as per the recommended maximum threshold of soil loss tolerance of 10 t ha−<sup>1</sup> yr−<sup>1</sup> for tropical areas [71]. Lake Bogoria-Baringo basin presented mean annual erosion rates >10 t ha−<sup>1</sup> yr−<sup>1</sup> for both periods while Lake Naivasha which recorded 10.8 t ha−<sup>1</sup> yr−<sup>1</sup> (Table 9) in year 2015. These values are in agreement with study by Mati et al. [24] which reported >10 t ha−<sup>1</sup> yr−<sup>1</sup> mean erosion rates in the Ewaso Ngi'ro basin. Flooding and the adverse effects of soil erosion within the Kenya GRV lakes has been attributed to the geomorphology of the lakes' environment and climatic factors [23]. Research is ongoing to find more definitive explanations for the recent rising levels of these Great Lakes [72,73]. Lakes Baringo and Naivasha are bordered mostly by flats lands while Lakes Bogoria and Nakuru are located in valleys enclosed by eastern and western rift escarpments [17]. Mubea and Menz [74,75] works revealed increasing urbanization and land degradation patterns in Nakuru district. This compounded with recent climatic changes [50] can offer some explanations why Lake Nakuru basin recorded the highest rise in soil loss rates among the Great Lakes. Rapid increment in floriculture and horticulture farming, over cultivation near river banks, and population growth are some of anthropogenic factors contributing to land degradation in the Lake Naivasha basin [76]. Willy et al. [77] survey study in Lake Naivasha basin showed low implementation of soil conservation practices with only 16% of sample households employing a combination of terracing, contouring, and grass strips. Lakes Baringo and Bogoria are located in Baringo County whose economy relies heavily on livestock which contributes 70% of its total income and supports 90% of its population [9]. Baringo County experiences frequent droughts; therefore, overgrazing can put pressure on land resulting to desert-like conditions [78]. Despite the highest mean annual precipitation and slope gradients for tree cover areas, forests presented lowest mean erosion rates for both periods (Table 5) emphasizing on the values of trees in soil erosion control. However, croplands had highest mean erosion rates indicating that intensive agricultural activities in areas with steep slopes significantly increase soil erosion threat within the Kenya GRV region. Fenta et al. [7] noted the high susceptibility to soil erosion of Kenyan highlands especially in disturbed forests or under sparse vegetation. Previous studies including [14,41] have also reported high erosion rates in highland areas with forestlands or poor vegetation due to deforestation, overgrazing, wildfires, and land cover changes. Koirala et al. [67] is in agreement with the concept that soil erosion increased proportionately with slope in mountainous regions while Schürz et al. [36] also reported high erosion rates in forested districts located on highlands e.g., West Pokot and Marakwet. In addition, recent study by Kogo et al. [31] in western Kenya (a subset of the KGRV region) revealed high rates around highlands e.g., the Mt. Elgon. Large bare areas in the ASAL which might potentially have high erosion rates recorded relatively low actual mean rates due to low rainfall and erosivity values. Significant bareland to grassland land cover conversions resulted in slight reduction of soil loss rates in ASAL areas, e.g., Turkana and Marsabit districts. The top three priority regions (Table 6) with erosion rates > 20 t ha−<sup>1</sup> yr−<sup>1</sup> contribute approximately 10% of the total erosive prone areas in the year 2015 and include highland districts located across the steep escarpment and ranges. Such areas are in urgent need of soil water conservation measures to mitigate heavy soil losses. Most of the protected areas are within forests or highland areas and recorded high erosion rates (Table A5) which is comparable with other estimated erosion rates for protected sites in other tropical lands as shown in References [28,29].

To assess the validity of RUSLE method for this region, the findings of the study were compared with areas of similar geo-environment and climatic conditions in the Eastern Rift Valley (EAR) region and seen to be analogous. For instance, our results coincide with those of previous studies by: Aneseyee et al. [79] evaluated the mean soil erosion rate in the neighboring Omo-Gibe Basin in the Ethiopian Rift Valley to be 17.65 t ha−<sup>1</sup> yr−1, Tamene et al. [71] found the mean soil loss rate of Laelaywkro catchment in Northern Ethiopia to be 20 t ha−<sup>1</sup> yr<sup>−</sup>1, Gizaw et al. [80] revealed that mean annual soil loss of Somodo watershed in South West Ethiopia is 18.69 t ha−<sup>1</sup> yr−<sup>1</sup> and Ligonja and Shrestha [81] reported a mean erosion rate of 15.7 t ha−<sup>1</sup> yr−<sup>1</sup> in Kondoa, Tanzania (Table 12). The districts' mean erosion rate values within the study area were consistent though not equal to the median and mean of soil losses values that resulted from the USLE model ensemble calculated by Schürz et al. [36]. In line with other studies in the East Africa region, terracing was found to be a highly effective soil erosion control measure especially for croplands located on the Kenya GRV highlands and the High Rainfall AEZ. Terracing and use of stone bunds was in practice across the Eastern Rift Valley (ERV) in small scale since the prehistoric periods as evidenced by ancient agricultural landscapes situated in the ASAL areas, i.e., Marakwet (Kenya), Engaruka (Tanzania) (Figure 10a), and Konso (Ethiopia) [82]. In Kenya, many of these terraces that had shown significant results were demolished or abandoned in retaliation to the colonial authority [20]. A sizeable number of smallholder farmers currently employ terracing and contouring within the Kenya GRV region (Figure 10b).

**Table 12.** Previous studies in and around the Great Rift Valley region of Kenya.


**Figure 10.** Soil conservation practices in the Eastern Rift Valley (ERV) region: (**a**) an archived aerial image (1960s) showing abandoned stone bunds and terraces of the semi-arid, historical agricultural landscape on the foot of the ERV escarpment at Engaruka, Tanzania; (**b**) Google Earth image taken on 26 July 2019 depicting soil conservation measures in Narok County, Kenya GRV.

Ruto et al. [86] revealed that terracing reduced soil erosion activity in Narok County and significantly increased maize and beans yields. Despite their effectiveness in controlling runoff, terraces, and stone bunds can be the source of erosion if poorly maintained or abandoned over time [64]. The high mean water erosion rates in the river basins can be attributed to negative land use land changes (Figure 2b) as well as neglect in adopting effective soil water conservation measures (Figure 10). Zhunusovaet al. [87] indicated that the single use of terraces had negative impact on crop yield in the Lake Naivasha basin while Reference [88] found that mulching and ground cover can be ineffective in controlling runoff flow on croplands with steep slopes as those in Kenya GRV region. In addition, Willy et al. [77] reported combined control measures (multiple soil conservation practices) can be adequate within the Lake Naivasha basin. Ten out of the thirteen water basins in the Kenya GRV region are located in ASAL regions characterized by lowland pasture, desert shrubs, exposed barren areas with sparse vegetation and poor land, and animal husbandry; thus, susceptible to water erosion (Figure 11). In Kerio Valley basin, increase in barelands, degraded forests coupled with recent high rainfall intensity increased soil loss resulting in heavy sedimentation as evidenced by very low water levels of Lake Kamnarok in the downstream areas [12]. Although the RUSLE method has been applied widely in different landscapes with significant results, its accuracy largely depends on the type of dataset (resolution, up-to-date, preference of primary over secondary data) and data manipulation methods [30]. The method suffers some limitations and its applicability in mountainous terrain remains doubtful [89]; possibly explaining the very high mean erosion rates recorded on escarpments and ranges within the Kenya GRV region.

**Figure 11.** Google Earth images showing spatial occurrence of water erosion in the Kenya GRV region: (**a**) image taken 15th May, 2017 showing exposed soil areas in Marigat District within the Lake Baringo basin; (**b**) image taken 15th May 2017 indicating poor soil conservation measures in Lorwak in Baringo County; (**c**,**d**) a time series images taken 9t October 2017 and 18th December 2017 respectively showing water erosion severity on a bare area in Narok County during the short rain period (October–December, 2017).

The soil erodibility factor for this study did not include pertinent factors, e.g., level of soil weathering, resistance against dispersion and crusting [90]. The mean annual erosion estimated in the current study also does not account for rainfall erositvity and vegetation seasonal variability. The Van der Knijff algorithm *C* = *exp* <sup>−</sup>*<sup>a</sup>* <sup>∗</sup> *NDV I <sup>β</sup>*−*NDV I* [91] where *a* and *β* are the parameters that determine the shape of the (Normalized Difference Vegetation Index) NDVI-C curve have been employed by several studies [28,29] in the East Africa (EA) region to estimate vegetation cover factor although it can lead to overestimation of *C* values in tropical regions with high rainfalls [30,41,92] (e.g., the forested areas of the KGRV). This equation gave high range of values for the KGRV (0.0–1.6) when compared to the values generated by the Durigon equation [92] *C* = (−*NDV I* + 1)/2 (0.05–0.65) recommended for tropical areas. This is after applying these equations on eleven MODIS (Moderate Resolution Imaging Spectroradiometer) NDVI datasets taken during the KGRV's wet seasons in the year 2015 for comparison purposes. Methods that assign *C* values to different LULC classes based on on-field determinations or combination techniques (e.g., involving image transformations and geostatistical analysis) are recommended for better estimation of C factor [4,30,36,71]. Table 13 shows the zonal statistics details between *C* factor map given by the Durigon equation on MODIS NDVI and the 2015 LULC map. The mean *C* factor values for forests are high (approximately 150 times higher than those used in this study) resulting to overestimation of the corresponding mean soil loss rates. This can give a false interpretation that croplands are better than forests in controlling water erosion in the KGRV.

**Table 13.** The mean *C* factor values for different land uses in KGRV and their corresponding soil loss rates for 2015.


Few works have been undertaken to estimate *P* factor values for the EA region [93] thus a maximum value of "one" was assigned to the Kenya GRV study area to indicate none conservation measures. The *P* factor method previously applied by other researchers in the region in which agricultural lands are assigned *P* values in relation to percent slope [41,46] (0–5%, *P* = 0.10; 5–10%, *P* = 0.12; 10–20%, *P* = 0.14; 20–30%, *P* = 0.19; 30–50%, *P* = 0.25; and 50–100%, *P* = 0.33) was noted to generate a comparatively low mean erosion rate (6.3 t ha−<sup>1</sup> yr<sup>−</sup>1) for croplands within the Kenya GRV after including it in the 2015 RUSLE model. Use of high resolution imagery with field data can be more sufficient [94]. In addition, generation of RUSLE parameters from published datasets with medium or coarse resolutions (data interpolations) may produce spatially variable erosion rates that can exceed acceptable tolerance levels [95]. These limitations notwithstanding, the RUSLE method was seen as a fast and practical approach in pinpointing potential erosion hotspots for such a vast region using limited data and the study results will be valuable in the management of the Kenya GRV region as well as provide useful guidelines in soil erosion investigations in tropical areas.

#### **5. Conclusions**

Water erosion is major source of soil degradation in Kenya whose land mass is dominated by arid and semi-arid lands. These ASALs are susceptible to natural hazards including soil erosion that can destroy vegetation cover resulting to land degradation hence

increase desertification risk. This poses a significant threat to agricultural production and food security in Kenya. The present study examined the magnitude of soil erosion rates using the RUSLE model in the Great Rift Valley region of Kenya due to its environmental diversity (i.e., combination of ASALs and agricultural lands) and important ecosystem services it provides in the country. The study is among the first attempts to quantify multi-temporal soil loss rates at the national scale for Kenya (with about 33% of the total land mass). Annual soil loss was found to be severe in the central and southern parts of region particularly along mountain fringes with high rainfall intensity for both years. The overall mean soil loss rate for the entire area fell under the tolerable erosion rate of 10 t ha−<sup>1</sup> yr−<sup>1</sup> in 1990 and 2015 although the substantial net changes in erosion rates for croplands underscores the need to devise effective anti-erosive interventions. Areas that require prioritized in soil conservation measures include more than half of all the protected areas as well as croplands in the central and southern regions of the Kenya GRV that presented mean erosion rates higher than 15 t ha−<sup>1</sup> yr−1. This also includes water basins: e.g., Lake Nakuru which has high urbanization trends as well ASAL basins, and Lake Bogoria-Baringo basin which has been adversely affected by human activities, e.g., agriculture with poor SWC measures and over reliance to pastoralism. Outcomes of this study can inform watershed managers on ways to reduce soil erosion rates, e.g., value of integrated conservation practices, curbing unregulated land use, overgrazing, limiting mass migration and deforestation as well as encouraging conservation tillage. The findings can further help policy makers plan for sustainable soil management strategies as the country gears towards achieving land degradation-neutrality.

**Supplementary Materials:** The RUSLE model factor maps for KGRV are available online at https://www.mdpi.com/2071-1050/13/2/844/s1.

**Author Contributions:** Formal analysis, methodology and investigation and writing; G.W., L.Y. and J.Z. Project supervision and administration; L.Y. and Y.N. Reviewing, statistical analysis and discussion; T.N., B.K., J.d.D.N. and Y.N. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Key Research and Development Project of China through grant agreement No 2020YFC1521900 and 2020YFC1521901. This work has also been supported by National Key Research and Development Program (No. 2014A8007007020) and China Scholarship Council 2014GXYB33 through Sino-Africa Joint Research Centre.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data available in a publicly accessible repository that that does not issue DOIs.

**Acknowledgments:** The authors would like to thank the editor and anonymous reviewers for their valuable comments and suggestions to improve the quality of this paper. The authors thank Yves Hategekimana, Fidele Kamarage and Felix Mutua for their valuable assistance.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A**

**Table A1.** Estimated Water-induced Mean soil loss rates under different Land Use and Land Cover (LULC) Conversions within the Great Rift Valley region of Kenya.




**Table A3.** Estimated mean soil loss and mean slope per district of erosion-prone areas experiencing Land Use/Land Cover Changes (LULCC) between 1990 and 2015.

**Table A4.** Distribution of estimated soil erosion rates under different soil types in the KGRV.



**Table A5.** Estimated Water-induced Mean soil loss rates per Protected Areas within the Great Rift Valley region of Kenya.


**Table A5.** *Cont.*

#### **References**


### *Article* **Climate-Smart Adaptations and Government Extension Partnerships for Sustainable Milpa Farming Systems in Mayan Communities of Southern Belize**

**Kristin Drexler**

**Citation:** Drexler, K. Climate-Smart Adaptations and Government Extension Partnerships for Sustainable Milpa Farming Systems in Mayan Communities of Southern Belize. *Sustainability* **2021**, *13*, 3040. https://doi.org/10.3390/su13063040

Academic Editors: Maurizio Tiepolo and Marc Rosen

Received: 25 January 2021 Accepted: 7 March 2021 Published: 10 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

American Public University System, Charles Town, WV 25414, USA; kristin.drexler@mycampus.apus.edu

**Abstract:** There are disproportionate adverse impacts related to climate change on rural subsistence farmers in southern Belize, Central America who depend directly on natural resources for their food and livelihood security. Promoting a more resilient farming system with key climate-smart agriculture (CSA) adaptations can improve productivity, sustainability, and food security for Mayan milpa farming communities. Once a sustainable system, the milpa has become less reliable in the last half century due to hydroclimatic changes (i.e., droughts, flooding, hurricanes), forest loss, soil degradation, and other factors. Using interviews with both milpa farmers and Extension officers in southern Belize. This qualitative study finds several socio-ecological system linkages of environmental, economic, socio-cultural, and adaptive technology factors, which influence the capacity for increasing CSA practices. Agriculture Extension, a government service of Belize, can facilitate effective CSA adaptations, specifically, an increase in mulching, soil nutrient enrichment, and soil cover, while working as partners within Maya farming traditions. These CSA practices can facilitate more equitable increases in crop production, milpa farm system sustainability, and resilience to climate change. However, there are several institutional and operational barriers in Extension which challenge their efficacy. Recommendations are presented in this study to reduce Extension barriers and promote an increase in CSA practices to positively influence food and livelihood security for milpa communities in southern Belize.

**Keywords:** climate-smart agriculture; socio-ecological systems; extension; Belize; milpa; food security; sustainability

### **1. Introduction**

For centuries, the traditional practice of milpa farming has been sustainable and reliable as the major food and livelihood source for Mayan milpa communities in southern Belize [1–4] as farmers allow areas to regenerate to a mosaic of forest succession stages and crop diversity [5–8]. In the last 50 years, however, the slash-and-burn aspect of the milpa has become less reliable and less sustainable due to environmental factors, such as hydroclimatic changes (i.e., droughts, flooding, hurricanes), forest and biodiversity loss, pests and crop disease, soil degradation and other factors in combination with socioeconomic and governance factors such as poverty, population growth, land tenure, and marginalization [8–16]. These factors have multiple systemic impacts to the resilience of milpa communities.

Government response and action is needed to promote climate-smart agriculture (CSA) practices and positively influence food and livelihood security in Belize. CSA practices can "increase productivity in an environmentally and socially sustainable way, to strengthen farmers' resilience to climate change, and to reduce agriculture's contribution to climate change" [17] p. 14. Government agricultural Extension service in Belize is in an effective position to promote CSA practices in Maya milpa communities because Extension works within the cultural traditions of the milpa system as partners in the process [10]. However, there are multiple barriers for Extension which challenge its efficacy, including

milpa farmers' land tenure, taxation, and poverty, and Extension's lack of operational budget, lack of technical training in CSA technologies, and a lack of staff [15]. Without a more effective Extension service facilitating CSA adaptations, there are implications for unsustainable crop production and food and livelihood insecurity in milpa communities of rural southern Belize [10,16–18]. The purpose of this study is to both examine three promising CSA practices—mulching, soil nutrient enrichment, and cover plants—and make policy recommendations to reduce Extension barriers to promote CSA practices for positive influences on food and livelihood security in southern Belize.

#### **2. Background**

Food and livelihood security for milpa farmers in southern Belize depends largely on Government response and promotion of climate-smart practices. Food security is the ability to provide present and future generations with a reliable food supply; it considers multiple factors and depends upon reliable crop production while sustaining a healthy ecological balance in a farming system [19–22]. Food security is dependent upon sustainable agriculture—the enhancement of crop production while sustaining a healthy ecological balance within agro-ecosystems [22]. Sustainable agriculture involves economic, environmental, social and other factors to promote food and livelihood security for communities [19,21].

#### *2.1. Forest Loss and Climate Change Vulnerability in Belize*

Rural communities in Belize are vulnerable to resource loss and degradation due to climate change, forest and biodiversity loss, and other factors [10,16–18]; these impacts, due, in part, to agro-industry and slash-and-burn agriculture, are exacerbated by rapid population growth, increased input of fertilizers, and farming on degraded soils [18,23]. There are implications for unsustainable agricultural systems and with that, the loss of forest, water availability, erosion control, and other needed natural resources and ecosystem services unless there is a strong government response in Belize. These ecosystem changes are "expected to threaten the sustainability of social, economic, and ecological systems" [24] p. 8.

Intact forests regulate climate, protect soils and water, and contain over 75 percent of global terrestrial biodiversity [25]. In Belize, the forest cover is roughly 60% but declining [26]. The growing rate of forest and biodiversity loss in Belize has compounding ecosystem pressures related to climate change, pollution, environmental degradation, and continual expansion of farms into forests [18,27,28]. Agriculture is the most significant anthropogenic driver of deforestation globally [11,25,29] and in Belize [12,26]; agriculture in the tropics directly impacts forest loss and is "responsible for nearly 85% of deforestation [and] 45% of deforestation in the humid tropics [is] due to shifting cultivation" [30], para. 17.

Large-scale climate and ecosystem changes in southern Belize have distinct impacts on the environment, crop production and economy, food security, public health, culture, and other factors in Belizean milpa communities [24,31–33]. Climate change impacts perceived by farmers in Belize include a lack of rain, increased heat and sun exposure, offset rainy seasons, increased storm intensity, and an increase in pests and crop diseases [32]. In addition, climate change accelerates soil erosion and land degradation. These factors negatively impact crop reliability, which is linked to livelihoods and resource security, community health, cultural traditions, and other factors [15,34]. There are disproportionate adverse impacts related to climate and ecosystem change on the rural poor, who depend directly on natural resources for their food and livelihood security [12,15,17,35,36]; these impacts perpetuate a cycle of environmental degradation, poverty, and vulnerability to climate and ecosystem changes [37].

#### *2.2. The Milpa Farming System in Belize*

A milpa is a small-scale shifting cultivation system of subsistence farming [4,38] traditionally involving slash-and-burn and/or slash-and-mulch practices [39,40]. Mulching and nutrient enrichment have also been a part of the traditional Maya milpa farming practice for centuries [40–42]. The milpa is a significant aspect of Maya culture and tradition as Maya identity, ceremony, community, and livelihood are all rooted in the milpa [9,43]. Milpa crop production is used for subsistence and selling at local markets [4,38]; milpas provide most of a family's need for food, wood, and income [11,44].

Milpa practices include clearing small areas of forest to plant a diversity of crops primarily corn, beans, and squash—on nutrient-rich soil [7,44]. Through crop diversity, the milpa can sustainably increase milpa productivity and promote food security and food sovereignty, the "right to healthy and culturally appropriate foods" [43], p. 396. However, the milpa system is "not indefinitely resilient, particularly in an era of global economic and environmental change" [12], p. 75. Specifically, the slash-and-burn aspect of traditional milpa farming (clearing and burning of small areas of forests for crop rotation) is no longer sustainable with changing climate conditions, increasing human population, and natural resource competition [38]. Milpa farmers who exclusively practice slash-andburn agriculture are more vulnerable to livelihood and food insecurity [10,12,17]. Burning reduces carbon stocks and the intense heat during burning can destroy critical root and seed banks [45]. Moreover, water-holding and nutrient status declines which dramatically increases "risks of accelerated erosion, water runoff, and crop failure in times of below normal rainfall" [46], p. 112.

#### *2.3. Managing Climate-Smart Agriculture (CSA) in Belize*

There is a need to manage resource loss and climate change vulnerability in Belize. Climate variability and extreme events (e.g., droughts, storms) are expected to become more frequent and damaging to water resources and agro-ecological systems in Belize in the coming decades [31]. The agriculture sector in Belize is especially vulnerable to climate change "not only due to its geo-physical location and hydro-meteorological hazards, but it is also due to the shortcomings of the current disaster risk reduction and response mechanisms to effectively mitigate the impacts" [31]. There is also a lack of institutional expertise to handle foreseeable climate change impacts [31].

The aim of climate-smart agriculture (CSA) is to "increase productivity in an environmentally and socially sustainable way, to strengthen farmers' resilience to climate change, and to reduce agriculture's contribution to climate change" [17], p. 14. CSA practices such as mulching and nutrient enrichment, along with improved land and water management, can result in higher and more stable yields, less production risk, increased system resilience to climate change, and lower greenhouse gas emissions. Therefore, CSA practices contribute to better food and livelihoods security for farming communities [34,47,48].

In Belize, the government vision for agriculture—the main engine of economic growth in Belize—incorporates several pillars including sustainable production and innovative technologies (Pillar 1), nutrition and food security, especially for rural populations (Pillar 3), and "climate change adaptation, environmentally sound production practices, conservation of natural resources, and risk management mechanisms" (Pillar 4) [49], p. 12. Continuing challenges to sustainable agriculture in Belize include: High and increasing poverty and unemployment, the rising inequality and access to food vulnerable populations, the steady decline in agriculture competitiveness, and increased scarcity of natural resources worsened by natural disasters and climate change [49]. These challenges are exacerbated by global financial instabilities which have negatively impacted employment, food and nutrition security, poverty, and inequity in Belize [49].

To improve the agricultural sustainability and reduce impacts of natural disasters and climate change, government action is needed, including the "promotion of more resilient farming systems and practices [e.g., climate-smart practices], as well as sound coordination, exchange of information, methodologies, and tools between experts and institutions" (31, para. 12). Increasing CSA practices can sustainably mitigate climate change impacts and support food security under a changing climate [15,50–52], while maintaining the health of ecosystems [20] and potential equitable increases in production in Belize. Otherwise, "marginal areas may become less suited for arable farming" [17], p. 15, resulting in less food and livelihood security for milpa farming communities in Belize.

#### 2.3.1. Mulching and Soil Cover for Water and Nutrient Holding

In Belize, climate-smart agriculture practices include mulching, soil nutrient enrichment, and soil cover; mulching avoids burning of debris and allows farmers to let organic matter decay on site. Mulching improves water holding capacity, soil organic matter (SOM), fertility, and stability, as well as reducing runoff and weed growth [12,42,47,51]. Further, mulching can improve soil water-holding by adding crop residues and manure to soil which effects soil properties and nutrient cycling, as well as lowering emissions [15,53]. Mulching has also been found to regulate surface temperatures, thus improving moisture and germination as well as other benefits for crop productivity [39,40,54].

Practiced by about half the milpa farmers in southern Belize' Toledo District [32], mulching has similar planting and harvesting timing and is beneficial because it restores degraded soils, provides shorter fallow periods, and stabilizes crop yields [39,40,54,55]. Also, mulching in addition to soil cover, such as mucuna beans can lead to higher yields due to decreased on-farm erosion and nutrient leaching, lower grain losses due to pests, and reduced labor for weeding and fertilizer application [47,56,57].

#### 2.3.2. Soil Nutrient Enrichment

Soil nutrient enrichment involves farming inputs that improve the soil conditions for production [39,40,54]. Soil enrichment practices can include adding chemical or nonchemical fertilizers and integrating effective microorganisms (EM) to break down slashed debris faster and build soil fertility [10,15]. There may be a need for farmers to purchase and use fertilization inputs in mulch systems, although that is debated in the literature. Two studies state fertilization inputs are essential to achieve good yields under fire-free conditions, although this cost may be offset from increased yields [15,58]. Other studies find that external fertilizer inputs were avoided with mulching where there was an increase in soil organic matter and water holding capacity [12,59].

There are some disadvantages to mulching and nutrient enrichment for farmers. Aside from the potential need to purchase fertilization inputs to enrich soil, mulching might also have a potential to increase snakes or animal vector encounters. Additionally, although mulching benefits are largely agreed to increase sustainability of yields, the increase in yield amount is debated due to slower nutrient release from the decomposing vegetation compared to burning [51]. Overall, mulching was found to have important farm system sustainability benefits, such as improving soil nutrients, regulating surface temperatures, improving moisture and germination, and increased crop productivity and sustainability [39,40,54].

#### *2.4. Government Agriculture Extension in Belize*

CSA practices and sustainable crop production depend upon government policies and action to shape agriculture adaptation response at the local level [17,32,49]. Government Extension services provide scientific knowledge and promote climate-smart agriculture practices, technologies, and innovations through farmer education and demonstrations [49,60,61]. Globally, Extension has a strong institutional expectation to inform, educate, and facilitate best practices for farmers [60] and to improve agriculture sustainability and promote "more resilient farming systems and practices" [31] para. 12 by working within the local sociocultural traditions. Smallholder farmers are able to adapt to environmental changes using their traditional knowledge and experience and by adopting climate-smart agriculture adaptions [62].

In Belize, agriculture Extension is in an effective position to promote climate-smart practices in Maya milpa communities because they can work within the cultural traditions of the milpa system as partners in the process [10]. In doing so, Extension can influence adaptive capacity by transferring and promoting CSA technologies and site-specific technologies, such as water management, cover plants, enrichment, and mulching [10,12]. These practices facilitate a more productive and resilient agriculture system [31] and build resilience in milpa communities [2,10,63,64].

With Extension support, farmer capacity can be improved to innovate, solve problems, and adopt CSA practices; however, this will require continuous facilitation, capacitybuilding and support over time [56,65]. There are multiple barriers for Extension efficacy in promoting CSA practices in southern Belize, including governance, land tenure, taxation, poverty, and other challenges [15,49]. Further, there are institutional barriers within Extension including a lack of operational budget, lack of technical training in CSA technologies, and a lack of staff allocated from the national Extension office [49]; presently, there are four Extension Officers in southern Belize' Toledo District who are responsible for a large rural district of 52 communities [32].

#### *2.5. Socio-Ecological Systems (SES) Framework*

As part of the socio-ecological system (SES), milpa communities experience system impacts and can be more vulnerable to ecosystem changes such as climate change. Socioecological systems (SES) is an effective framework to study climate-smart agricultural adaptation in milpa communities in Belize [10,17,66] as SES is complex, systemic, cumulative, and intertwined with human systems [67]. This study examines perceptions of climate-smart practices from milpa farmers and agricultural Extension officers in southern Belize using a SES framework. SES is a flexible framework which considers the interrelationships, linkages, and synergies between multiple trans-disciplinary factors—social, economic, environmental, cultural, governance, justice and other factors; SES also involves inclusion and community-based partnerships and adaptive management [68–70]. Milpa farmers and Extension Officers in southern Belize can become more enabled partners in climate-smart solution finding. The socio-ecological system of milpa communities is a linked network where an impact on one part of the system—the loss or degradation of soil due to storm erosion, for example—can affect the human system, such as food security and farmer livelihoods [11,67,71].

#### **3. Methods**

This qualitative study uses Phenomenology and face-to-face interviews to examine common-lived experiences of climate-smart practices from the perceptions of milpa farmers and Extension officers. Phenomenology is well-suited for this study as it is both a philosophy and an inquiry strategy used to "develop an understanding of complex issues that may not be immediately implicit in surface responses" [72], p. 301. Phenomenology is useful to "investigate the relationship between participatory Extension methods and farmers changing to more sustainable practices" [73], p. 22.

The semi-structured interviews were both "purposive and prescribed from the start" [72], p. 302 and allow flexibility to ask participants deeper follow-up questions [32], in order to hear their stories, and to see the emergence of common experiences or phenomena through the participant's own words and descriptions [74–76]. During interviews, participants were asked both demographic questions and open-ended questions on topics such as milpa farming practices, socio-ecological system linkages to ecological changes (i.e., forest loss, climate change), barriers and conduits of sustainable agriculture practices, and other topics. Using Phenomenology, specific patterns, categories, and themes emerged from the interview data collected and analyzed [77–79]. New Mexico State University Institutional Review Board (IRB) approved all study protocols and interview questions; all interviews followed a voluntary and informed consent procedure.

#### *3.1. Setting of the Study*

In the southern Toledo District of Belize, three Extension officers and five milpa farmers from Pueblo Viejo and Indian Creek villages were interviewed for this study. Toledo District is the southernmost district in Belize; its population is nearly 50% Q'eqchi' (Kekchi) Maya, 20% Mestizo, and 17% Mopan Maya. There are also Garifuna, Creole, East Indian, and Mennonite populations [80]. Milpa households in each village were selected using a stratified random design. The sample subpopulation of 'primary (head) milpa farmer' for each selected household was intentional to elicit the perspective of farmers who have the most direct knowledge of local forests, soils, and agriculture systems. Two participants selected were Maya cultural and political leaders in their villages who spoke to the importance of the milpa as part of their cultural practice. Interviews of Extension officers were conducted in both office and field settings; three (of the four total) Extension officers in the Toledo District in southern Belize were interviewed.

#### *3.2. Data Analysis*

Using the multi-perspectival framework of Socio-ecological Systems (SES), a combination of processes was used in the data analysis, including 1. Open (analytical), 2. Axial (reduction and clustering of categories), and 3. Selective coding [77,81–86]. These processes involved creating categories of like taxonomies and assembling structures or groups of themes into conceptual diagrams to show relationships and linkages [82]. Selective coding is a form of data synthesis where the intersection or integration of emergent thematic categories are first "crystallized" [10]. Crystallization uses multiple perspectives to blend data to produce thick description and knowledge of a phenomenon as well as a deepened, inclusive, multi-perspectival, and complex interpretation of it [85,86]. Selective coding results as the intersection of the main categories and themes [77] where categories are systematically related or conceptually linked in a multi-perspectival and holistic way [86]. From this data analysis and synthesis, dominate themes emerged and were categorized in the following Results section.

#### **4. Results**

Through socio-ecological system (SES) examination of interview data from milpa farmers and Extension officers, this study finds direct and indirect influences of climatesmart agriculture (CSA) practices—specifically mulching, soil enrichment practices, and ground and soil cover methods—on milpa farming sustainability. Table 1 summarizes results from interviews with milpa farmers and Extension Officers. Participants in this study perceived direct and indirect, (a) environmental, (b) economic, (c) socio-cultural influences, and (d) adaptive technology potential from CSA practices on the sustainability and resiliency of Belizean milpa communities.

**Table 1.** Summary of Results: Perceived Environmental, Economic, and Socio-cultural influences, and Adaptive Technology potential from climate-smart agriculture (CSA) practices (mulching, soil enrichment, and ground cover) on milpa sustainability in Belize.


#### *4.1. Environmental Influences*

Milpa farmers and Extension officers perceived positive environmental influences from CSA practices on milpa sustainability. For the purposes of this article, environmental influences, include impacts to air quality, soil nutrients, soil water holding, land erosion, and pests and disease. This study finds CSA practices of mulching, soil nutrient enrichment, and soil cover have positive environmental influences for increasing soil nutrients, ground cover and moisture, controlling erosion, and managing crop pests and disease. Extension officers interviewed for this study perceived both mulching and nutrient enrichment to be climate-smart and beneficial for milpa famers.

#### 4.1.1. Benefits from Mulching (vs. Burning)

Milpa farmers and Extension officers perceived less impact from mulching over burning. All milpa farmers interviewed for this study practice slash-and-burn; one farmer explained his process: "I will soon start to chop bush, and then it dries, and then [I] burn it, and then plant it. You chop more bush to plant more [crops]." Some also practice mulching; one milpa farmer described his preference for mulching over burning due to the benefit of less erosion: "[We] just leave [debris] there and it'll get rotten, right? Leave the stump right there, because the stump—it holds a lot of soil [and] when it's raining, it won't flush off. So, just leave the stump right there until it gets rotten".

Extension Officers interviewed explained mulching provides effective ground cover and erosion control; also mulching keeps more moisture and fertility in the soil. Whereas, burning exposes and heats up the soil, causing nutrient loss. An Extension officer explained the benefits of leaving the vegetation to rot in the mulching process: "[The grass] covers the soil [and] ... there's a little moisture by the roots of the plant [and] it will keep the soil cool instead of in the hot sun ... so it does work. It does work." One Officer explained burning and not leaving debris on the soil causes erosion: " ... then you have a long drought [and then] how do you keep moisture? And, those are the things that we have to make farmers aware of—it's a chain of reaction." There is a disadvantage to mulching in "that it's too bushy and people don't want to go in there ... because it attracts maybe snakes and other things" [32]. However, there was an overall positive environmental influence of mulching, nutrient enrichment, and soil cover on milpa sustainability.

Extension Officers see other CSA benefits with mulching, including reduced air emissions, better water management, and the use of non-chemical inputs for crop pests and disease. Climate change creates the condition for unreliable water and a higher incidence of crop pests and disease; one Extension Officer described how this affects crop production: "A high incidence of pests (are) noticeable now ... so, all of these things—and a limited water supply—all of these things are affecting agriculture, in general ... and those things limits our work as well." He stated that years ago, they did not have to think about the climate, but now they have to "be [climate] smart." An Extension officer stated they "need to do a little bit more public awareness in terms of the negative effects [of burning]" due to air pollution, global warming, and other effects.

#### 4.1.2. Benefits of Nutrient Enrichment and Soil Cover

Milpa farmers traditionally rotate crops on nutrient-rich "black" soil due to the nutrient depletion in farmed soil over time [10]. One farmer stated if there were soil enrichment assistance for farmers—specifically information and financial assistance for inputs—he would not need to "chop" forest to use the enriched soil. Extension Officers want to demonstrate to farmers that climate-smart alternatives (i.e., mulching and enrichment) work. Increasing nutrient enrichment such as effective microorganisms (EM) and using nitrogenfixing cover plants can benefit milpa farmers. One Officer educates farmers and promotes EM. He stated, "A lot of farmers, they are starting to use organic material—meaning chicken manure. They are using a lot of EM agriculture to build up the soil fertility." The same officer also explained the benefits of mucuna beans for nutrient enrichment:

We have some farmers that benefit from the training as well, because, at some point, we introduce some types of fertilizer that you incorporate in the soil ... [for example] mucuna beans: The Mennonites [presuming he means the less mechanized Amish community] use it a lot, you know; they don't use a lot of synthetic fertilizer, they only use these types of mucuna beans.

Another Extension officer promotes arachis (*Arachis glabrata*), a wild peanut perennial. Arachis is useful for milpa farmers as an effective ground and soil cover and as a nitrogenfixing plant. These climate-smart practices mimic or replicate the nutrient cycling in forest ecosystems while allowing for sustainable production of agriculture [46].

#### *4.2. Economic Influences*

Milpa farmers and Extension officers perceived mostly positive and some negative economic influences from CSA practices on milpa sustainability. For the purposes of this article, economic influences include impacts to farmer income, farmer expenses, farmer time, Extension expenses, and impacts to alternative income (tourism). This study finds climatesmart practices, particularly mulching, soil nutrient enrichment, and soil cover, have overall positive economic influences for increasing farmer income and reducing farmer expenses. However, Extension's limited budget to promote CSA practices is a barrier.

From an economic perspective, milpa farmers are concerned about maintaining production (for subsistence and selling at local markets) and reducing costs and expenses related primarily to controlling pests and disease and adding fertilizer inputs—in their daily farming practice. Some farmers interviewed worried about increasing pests resulting from climate warming; their lack of knowledge of pests and pesticides management was a top concern [32]. Most often in milpa communities, pest management includes the use of chemical fertilizers and pesticides, which are an additional cost for farmers [32].

#### 4.2.1. Soil Enrichment Cost-Benefit

Soil enrichment involving fertilizer inputs (chemical or nonchemical) is a cost to farmers; all farmers interviewed for this study stated they buy and/or use fertilizer inputs. Although an upfront cost, one study concluded that using fertilizers increases farmer yields enough to compensate for fertilizer costs [58]. Moreover, adding nonchemical enrichment or soil cover (i.e., arachis) can be low to no cost. Milpa farmers who rotate crops in cleared forest areas avoid fertilizer cost by using nutrient-rich black soil. One farmer explained that otherwise, the soil gets too dry and hard; "but, if we change every year, it doesn't need fertilizer. Yah, just normal planting—organic ... That's why we maintain for we [sic] forest." Conversely, another farmer explained that keeping forests intact is important for his village's economic development and tourism industry:

We understand the slash and burn is [bad]—sometimes for humans, for us and also for a wildlife—and, so, we are trying to avoid that now. We are working very closely with the village leaders [to develop potential tourism in the village] ... because we need to take care of our forest, including creeks, rivers, and streams, and so forth.

#### 4.2.2. Extension Barriers

Climate change negatively impacts farmer food security and livelihood which can impact Extension efficacy in promoting and facilitating sustainable agriculture practices in milpa farming communities [32]. However, there are institutional and operational barriers in Extension services, including a lack of government funding for daily operating, a lack of technical training in CSA technologies, and a lack of staff. One Extension officer noted: "We need support from [the national office] because we cannot do it alone ... We need to prioritize [climate-smart] topics because everything now is climate change ... everything is focused around climate change and resilience." To cope with the low numbers of staff, Extension officers stated they have to collaborate and share resources with other government agencies and nongovernmental organizations to carry out some aspects of their Extension duties (i.e., sharing vehicles to reach farmers in remote communities).

#### *4.3. Socio-Cultural Influences*

Milpa farmers and Extension officers perceived both positive and negative sociocultural influences from CSA practices on milpa sustainability. For the purposes of this article, socio-cultural influences are defined as impacts to farmer traditions and heritage, adaptation responses, adopting new technologies (i.e., effective microorganisms), and community resilience. This study finds climate-smart practices, particularly mulching, soil nutrient enrichment, and soil cover have overall positive socio-cultural influences for

maintaining milpa socio-cultural practices and traditions. However, there are barriers for some farmers in adopting newer CSA technologies and practices.

#### 4.3.1. Climate Change Impacts to Milpa Culture

Milpa farmers and Extension staff perceived direct impacts from climate change more intense sun/heat, lack of rain, offset rainy season, increase pests and disease [32]; in turn, these impacts effect the cultural traditions of milpa communities and their ability to provide corn and beans for subsistence. One farmer stated: "I am waiting for that rain because I have some corn that is (small) [demonstrates small size] that haven't gotten water for a good while—so, I'm hoping and wishing that this week it will soon rain." Another stated: "I tried to plant vegetables but he [sic] no grow—he dead. I planted, but he neva (never) grow good. I don't know why, because maybe he got sun too much;" and "the rains are different than what they were because, it doesn't rain too much ... I am worried about crops—worried about the weather." A third farmer described the offset rainy season: When they expect the rainy season, there is strong sun and heat; they don't know "which time is for the right time" to plant. He explained: "We need to study the climate changes and the temperature (so we can) try to manage." This farmer also explained the exclusive practice of slash-and-burn, a traditional form of milpa agriculture, may not be culturally sustainable:

The only way we could damage [the milpa farming culture] for us is if we continue to slash and burn, and burn, and slash and burn—and, we believe that one day our crop will never come out good again because the fertile[ity] of the ground is washed off, so everything goes in the creeks, in the river; and, the land becomes poor and poor and poor and poor—and, so, now, we don't want to practice that because we understand the situation there. So, we believe that to maintain the soil, to treat the soil in a proper way ... not to cut down the trees or not to burn it—even though if you want to fall something—but, leave it there—just, leave it there, and it'll get rotten.

#### 4.3.2. Extension Barriers

Extension Officers interviewed for this study stated that climate change impacts to milpa farmers creates a barrier for Extension service efficacy. One Officer stated climate change is a real factor now: "A high incidence of pests (are) noticeable now ... so, all of these things—and a limited water limited water supply—all of these things are affecting agriculture, in general ... and those things limit our work as well." Another Extension Officer added: "Climate change is very, very important because all these pests and disease ... in a hot climate or little water available—always climate change issue is very critical now and there . . . and we make [farmers] aware of that".

#### 4.3.3. Cultural Adaptations and Adaptive Technologies

Increasing CSA practices such as soil nutrient enrichment can help milpa farmers plant on soil which mimics rich black soil (i.e., converted from cleared forest). However, some farmers perceive this socio-cultural adaptation is not needed; one farmer stated he prefers to clear forest because "black soil is better [to farm]" and that "works for us." Two farmers interviewed were interested in learning new technologies and adapting their practice; one stated an interest in intercropping and effective microorganisms (EM) for soil enrichment:

It would be interesting to bring something with the soil and mix it up—and put plants there like tomatoes. You could plant when you mix up the soil ... the [plants] come very good. And, with corn too ... Yes, yes—that would be interesting ... interesting. You bring some soil, you just mix it up, and plant some there.

Extension officers are promoting milpa farmers increase climate-smart practices, including slash-and-mulch farming and soil nutrient enrichment, as they are beneficial and sustainable milpa practices [32] which can increase farmer production and income. One Officer stated about half of the farmers in the district practice mulching; however, there

is a need for more public awareness of the negative effects of slash-and-burn practices by making it a priority to show farmers proof of workable CSA practices:

We need to continue to educate the farmers, right(?) ... because whenever you do your field visit and so on, you can see that some farmers, yes, they do adaptive technology very slowly; others, they go a little bit faster ... but some of them it's very hard. So, what I have found is that (we should) continue capacity-building, education ... (and) at some point, of course, especially the young farmers that are coming up—try to teach them the right way of doing agriculture—sustainably.

Another Extension officer is trying to educate and promote nutrient enrichment technologies such as effective microorganisms (EM), mucuna beans, and arachis: "A lot of farmers, they are starting to use organic material (such as) chicken manure, right? They are using a lot of EM agriculture to build up the soil fertility." He explained the benefits of mucuna beans:

We have some farmers that benefit from our training ... we introduce some types of fertilizer that you incorporate in the soil ... (for example) mucuna beans: the Mennonites [presuming meaning the less mechanized Amish community] use it a lot, you know; they don't use a lot of synthetic fertilizer, they only use these types of mucuna beans.

He also explained the benefits of arachis, a wild peanut perennial; arachis can be used for chicken feed in a mobile chicken coop. He explained arachis is an excellent ground and soil cover and as a nitrogen-fixing plant.

#### 4.3.4. Extension Partnership

Extension Officers perceived milpa farming practices are sustainable due to cultural traditions and knowledge passed down through generations. However, the Officers interviewed stated milpa traditions will only stay sustainable if farmers adapt to include CSA practices. An Extension supervisor in the national office stated district Extension officers can work within the cultural traditions of the milpa system to promote sustainable practices:

[We need] a way to demonstrate to [the farmer] a way to adequately compensate for what they are moving ... we need to look at injecting proportionate technology in the milpa system, and then look at how the farmers react to that injection. It's a learning process, not to challenge traditional [farming methods, but try to promote] a few [effective] agricultural practices like soil conservation, irrigation systems, and integrated pest management [87].

With time and resources, Extension Officers are in an effective position to promote sustainable agriculture production because they are partners with milpa farmers in both maintaining traditional milpa practices and adopting more sustainable CSA practices [15].

#### **5. Discussion**

The purpose of this study is to both (1) examine three promising climate-smart agriculture (CSA) practices—mulching, soil nutrient enrichment, and cover plants—and (2) make policy recommendations to reduce Extension barriers to promote the CSA practices for positive influences on food and livelihood security in southern Belize. From interviews with milpa farmers and Extension officers in Belize, this study finds CSA practices were perceived to have overall positive socio-ecological system influences on Maya milpa farming communities, including: (a) Economic, (b) Environmental, and (c) Socio-cultural influences, as well as (d) Adaptive technology potential; these influences were perceived as conduits for sustainable milpa agriculture. A flow diagram or conceptual impact model (Figure 1) of the influence categories was constructed (via PowerPoint software) using this study's perception data gathered during interviews. The model shows relationships and linkages between the SES influences as they relate to CSA practices.

*Sustainability* **2021**, *13*, 3040

**Figure 1.** A Socio-ecological Systems impact model from this study's climate-smart agriculture (CSA) influences environmental, economic, socio-cultural, and adaptive technology (potential)—on milpa farming sustainability and resilience in southern Belize.

Figure 1 is based on the conceptual framework by SES architect Dr. Elinor Ostrom, widely considered to be the foremost researcher on SES; her model shows a multilevel and multi-perspectival examination of SES factors, drivers, interactions, and outcomes with implications for adaptive management, multiple stakeholder coordination, collective action, and community-level application [68–70]. Other similar models informing Figure 1 which demonstrate complex, multi-perspectival linkages in a socio-ecological system include Parrot, et al. [70], Flora and Flora [88] community capitals, Cote and Nightingale [89], Gonzales, et al. [90], and Tenza, et al. [91]; these works show an emphasis on feedback dynamics where human systems can shape ecological components and vice versa.

The data analysis processes of selective coding and crystallization described in Section 2 created a complex and multi-perspectival interpretation where dominant themes emerged. The major influence categories were systematically linked [77,86] and show multiple intersections which directly and indirectly relate to CSA practices in milpa communities. In Figure 1, Environmental influences are represented by green boxes, Economic influences are burgundy, Socio-cultural influences are gold, and Adaptive technology potential are blue.

Figure 1 demonstrates linkages that can be useful to inform government policy and action. Intersections of linkages can be used to help identify priority areas which may have implications on the larger socio-ecological system. For example, the model demonstrates how soil enrichment practices are linked to potential adaptive technologies such as mucuna beans, arachis, and effective microorganisms; those are linked to less chemical inputs, a reduced need for costly fertilizer inputs, more soil stability by keeping forests intact (i.e., not needing to cut forests for black soil), maintaining Maya socio-cultural traditions, and so on altogether demonstrating a positive influence on the larger milpa socio-ecological system. Therefore, one Extension intervention (i.e., facilitating increased use of mucuna beans) can have positive impacts to other parts of the milpa socio-ecological system. Overall, the model shows increasing CSA practices of mulching, soil enrichment, and cover plants can foster higher crop production with more resource stability; the implications for this can mitigate disproportionate climate change impacts related to poverty, climate justice, resilience, and food and livelihood insecurity in milpa communities.

Figure 1 is intended to be a small picture of an otherwise larger and more complex milpa agroecological system with multi-dimensional, dynamic, non-linear (circular) feedbacks and flows [32]. Understanding these system relationships—and how each factor functions in the complex whole of the SES—is important as each decision a farmer makes to adopt CSA practices can advance the entire milpa agriculture system further [92,93]. Elements of Figure 1, specifically related to Extension and government action, were used to inform the policy recommendations in this study.

#### **6. Conclusions and Policy Recommendations**

From interviews with milpa farmers and Extension officers in Belize, this study finds climate-smart agriculture (CSA) practices of mulching, soil nutrient enrichment, and soil cover can have overall positive socio-ecological system (SES) influences on milpa sustainability in southern Maya Belize communities. Although milpa farming has been sustainable for centuries, global climate change and other factors such as poverty, population growth, forest loss, and land degradation have made the practice less so over the last 50 years. Promoting the increase of CSA practices on milpa farms has overall positive (a) environmental, (b) economic, and (c) socio-cultural influences and (d) adaptive technology potential on milpa sustainability, food and livelihood security, and resilience. By examining the perceived SES influences of CSA practices, this study makes policy recommendations to reduce government Extension barriers to promote the CSA practices for positive influences on food and livelihood security in milpa communities of southern Belize.

Promoting CSA practices necessitates Government involvement and action. Agriculture Extension in Belize is in an effective position to facilitate an increase in CSA practices because it has a strong institutional expectation to inform, educate, and demonstrate best practices to the public. Working within milpa cultural traditions, Extension Officers can promote an increase in climate-smart practices, while including milpa farmers as partners in the process. Specifically, promoting the practices of mulching, soil nutrient enrichment, and soil cover can have positive socio-ecological system influences and potential equitable increases in crop productivity. Government agriculture Extension services are needed to promote CSA practices in the milpa communities they serve. Recommendations here are targeted to Government Extension services in Belize at the national and district levels:

1. Increase District operational funds (i.e., vehicles, fuel), technical training, and the number of trained officers to enable Extension to promote farmer adaptation such as mulching, soil nutrient enrichment, and other nonchemical technologies (e.g., effective microorganisms, mucuna beans);


Incorporating these recommendations while continuing to work within the cultural traditions of milpa farmers as farmers as partners in the process, Extension services can promote CSA practices for a more sustainable milpa farming system for food and livelihood security in southern Belize.

**Funding:** This research received no external funding; it was part of Drexler's doctoral dissertation research (Drexler, 2019).

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of New Mexico State University; submission 17466|Extension Leadership and Sustainable Agriculture in Belize Forest Farming Communities: A Socio-ecological Systems Approach (dissertation); approved 2018-12-06 11:36:23.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the author. The data are not publicly available due to confidentiality and consent protocols related to audio recording of interviews; coding and categorizing of some transcribed responses is available.

**Acknowledgments:** We thank the Director of Extension and the Toledo District milpa farmers and Extension Officers for their time contributing interviews to this study.

**Conflicts of Interest:** The author declares no conflict of interest. Having no external funding source, there was no outside role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, nor in the decision to publish the results.

#### **References**


### *Article* **Impact of Climate Change Adaptation on Household Food Security in Nigeria—A Difference-in-Difference Approach**

**Oyinlola Rafiat Ogunpaimo 1,2,\* , Zainab Oyetunde-Usman <sup>3</sup> and Jolaosho Surajudeen <sup>2</sup>**


**Abstract:** Studies have shown that climate change adaptation options (CCA) are implemented to buffer the unfavorable climatic changes in Nigeria causing a decline in food security. Against the background of measuring the impact of CCA options using cross-sectional data, this study assessed how CCA had affected food security using panel data on farming households from 2010– 2016 obtained from Nigerian General Household Survey (GHS). Data were analyzed using the Panel probit model (PPM), Propensity Score Matching (PSM), and Difference-in-Difference (DID) regression. PPM showed that the probability of adopting CCA options increased with farm size (*p* < 0.01), extension contact (*p* < 0.01), and marital status (*p* < 0.01), but decreased with the age of the household head (*p* < 0.01). Credit facilities (*p* < 0.05), ownership of farmland (*p* < 0.01), household size (*p* < 0.01), years of schooling (*p* < 0.01), household asset (*p* < 0.01), and location (*p* < 0.05) also had a significant but mixed effect on CCA choices. PSM revealed that farming households that adopted CCA strategies had 9% higher food security levels than non-adopters. Furthermore, the result of the DID model revealed a significant positive effect of CCA on household food security (β = 5.93, *p* < 0.01). It was recommended that education and provision of quality advisory services to farmers is crucial to foster the implementation of CCA options.

**Keywords:** developing countries; welfare; panel probit model; adoption; propensity score matching

#### **1. Introduction**

The agricultural practices in African nations especially Nigeria largely rely on the natural weather conditions of the locality. Changes in the climatic condition of the country are evident in increased desert encroachment and extreme droughts in the Northern region [1,2], likewise the problem of persistent flood and erosion occurrence in the Southern region. Climatic variability and changes have been linked to erosion, increased flooding, environmental degradation [3], and a decrease in agricultural productivity [4,5].

Frequent and intense weather events as a result of climate change are likely to impact the welfare and food security status of both the rural and urban populace through poor food production, poor land availability, and reduced opportunities [6]. The optimal usage of land for crop and animal production, biodiversity restoration, health, and well-being can also be negatively impacted by increased temperature and precipitation changes, and increased weather fluctuations [7–9].

In accordance with the 2020 global food security index, Nigeria's food insecurity status is considered serious in the severity chart [10]. The Federal Ministry of Agriculture of Nigeria in 2014 estimated that 65% of the population is food insecure despite having more than half of all employments dependent on agriculture [5]. Among several other factors, heightened food insecurity among farm households is caused by limited access to

**Citation:** Ogunpaimo, O.R.; Oyetunde-Usman, Z.; Surajudeen, J. Impact of Climate Change Adaptation on Household Food Security in Nigeria—A Difference-in-Difference Approach. *Sustainability* **2021**, *13*, 1444. https://doi.org/10.3390/su13031444

Academic Editor: Maurizio Tiepolo Received: 29 December 2020 Accepted: 26 January 2021 Published: 29 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

credit, poor storage and improved agricultural facilities, and negative environmental influences such as erosions and floods [5]. Other reasons include the lower household income necessary for food purchases needed to attain food security [11], increased population growth [12,13], and a huge reliance on imported food items [14]. The 2030 United Nations Sustainable Development Goals are new global policies with the objective to restructure regional and national development plans over the next 10 years. The global policy aims to put an end to poverty and hunger, food insecurity, sustaining natural resources and the environment, and promote food and agriculture sustainability [15].

The International Symposium on Climate and Food Security (ISCFS) and Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) also recognized three critical global problems of poverty and hunger, increased population growth, and unfavorable weather and climate [16]. Agricultural production and climate change exhibit a feedback relationship, while agricultural activities may result in increased emissions and pollutions leading to climatic changes, climate change also influences agricultural output. Research has indicated that by 2030, the negative consequence of climate change on agriculture will be more severe across all the countries of the world [17]. Climatic changes already have an antagonistic influence on food security with the number of chronically undernourished people in the world estimated to have increased by 38 million in 2016, in addition to 777 million recorded in 2015 [18], thus, without proper implementation of adaptation and mitigation measures, climate variability and change will threaten the achievements of the SDG goals in eradicating poverty and hunger [17].

As stated by the United Nations Framework Convention on Climate Change (UN-FCCC, 2007) AR4 and buttressed by Field [19] and Steynor and Pasquini [20], Africa is predicted to be the continent most susceptible to climate change and variability; adverse climatic impacts are worsening the livelihoods, welfare, and regional and household food security in tropical developing regions [17]. Consensus exists that climate change and variability will have a significantly negative impact on all these aspects of food security in Africa [21]. Like other African countries, studies such as Ebele and Emodi [22] reported that the projected impact of climate change on West Africa's agricultural productivity could lead to a 4% reduction in the region's GDP; in Nigeria, the adverse influence of climate change on food security is evidenced by the changes in plant duration and output of cereal crops, reductions in aquatic life [23], and livestock failure [24]. In tackling climate change and variability, studies [25,26] indicate that CCA can promote household food security. While research had been conducted on factors influencing the choice of CCA by farming households and the impact of CCA options on household food security, very few studies if any had applied a panel data analysis approach to investigate the influence of CCA options on household food security in Nigeria.

Morland [27] stated that small scale farmers in Africa experience weather variability and other climate-change-related events. Among farm households, varieties of climateadaptation methods abound and this includes diversification and crop rotation, engaging in non-agricultural income-generating activities, practicing soil and water conservation techniques, adjusting the times they sow their lots, use of irrigation and creation of flood barriers, adopting improved seed varieties and fertilizers, tree-planting, and integrating crop production with livestock.

Adaptation to climate change and variability means anticipating the adverse effects of climate variability and taking appropriate action to prevent and or minimise the damage they cause or taking advantage of opportunities that may arise [28]. It involves the use of climate sustainable practices to reduce the negative consequences of climate change.

In many African countries, especially Nigeria, access to food will be severely affected by climate change. Africa is the region where climate change and variability have had the biggest impact on acute food insecurity and malnutrition, affecting 59 million people in 24 countries [7]. Given these discouraging prospects, it is no surprise that the adaptation strategies are vital to support the climate effects on food security. These strategies can indeed buffer against climate variability and play a crucial role in promoting the food security of farm households, thereby reducing the negative effect of climate variability on food security.

Thus, this research sought to investigate the impact of climate change adaptation on farm households' food security status by providing answers to the following questions: (1) What are the factors affecting climate adaptation strategies employed by farm households in Nigeria? (2) What is the impact of climate change adaptation on household food security?

#### **2. Materials and Methods**

This section describes the study area, the data source, a summary of farm household characteristics, the description of variables, and the method of data analysis.

#### *2.1. Study Area*

The study area is Nigeria situated 4◦ and 14◦ N, and longitudes 2◦ and 15◦ E. Nigeria is located in West Africa with its capital at Abuja created in 1976 having a total area of 923,769 km2 (356,669 sq mi) [29], making it the world's 32nd largest country. Nigeria is bordered to the north by Niger, to the east by Chad and Cameroon, to the south by the Gulf of Guinea of the Atlantic Ocean, and the west by Benin [29].

Like most countries in sub-Saharan Africa, the climatic condition in Nigeria is tropical with varying rainy and dry seasons. The Nigeria vegetation belts, also known as the agroecological zones range from the mangrove, freshwater swamps, and tropical rainforest which extends from the South-South, South-East to South-West regions. The tropical savanna grasslands zone is dominant in the middle belts with Sudan and Sahel Savanna in the Northern regions. Agricultural production is dominant in the northern agroecological zones, however, climatic changes and human activities such as continuous cropping, overgrazing, and bush burning especially in densely populated areas have highly impacted the agricultural vegetation.

Specifically, in the far northern areas, the nearly total disappearance of plant life has facilitated a gradual southward advance of the Sahara [29]. The map of Nigeria showing the different regions and states is illustrated as shown in Figure 1.

#### *2.2. Data Source*

Panel data on relevant socioeconomic, demographic, consumption, and production data of farming households from 2010–2016 were obtained from the World Bank and the Nigeria Bureau of Statistics (NBS) General Household Survey (GHS). The GHS is a nationally representative survey with respondents obtained from the 36 states of Nigeria including the Federal Capital Territory. Three waves of the Living Standard Measurement Survey (First wave—2010/2011 Ref. NGA\_2010\_GHSP-W1\_v03\_M [30], Second wave— 2012/2012 Ref. NGA\_2012\_GHSP-W2\_v02\_M [31] and the Third-wave—2015/2016 Ref. NGA\_2015\_GHSP-W3\_v02\_M [32]) was employed such that there are six-period panel data for the farm households from which 3500 farm households were used for this study.

#### *2.3. Summary of Characteristics of Farm Households in the Study Area*

The summary of farm households' characteristics is shown in Table 1. Age is an important determinant of farm activities. It is believed that younger people commit more energy to production activities, while older farmers are likely to be more experienced. The mean age of respondents was 50.4 and 53.2 years in 2010 and 2016 respectively. In relation to the gender of the household head, the majority of family households were male-headed. Male household heads accounted for 88.36% in 2010 to a slight reduction of 84.14% in 2016. This report buttresses the dominance of males in farming activities in Nigeria as reported by [33]. However, the decline in the proportion of male-headed households in the study area may be due to increasing awareness of female empowerment and capacity building in the area.


**Table 1.** Summary of Farm Households' Characteristics in the Study Area.

Author's Computation from LSMS Data 2010:2016.

As illustrated in Table 1, the majority (85%) of farming households in 2010 were married, this came down to 79.57% in 2016. In terms of educational level, the average years of schooling were 6 years with no significant change in 2016 implying that although farm households have access to formal education, the level of education among farming households in Nigeria is still low. However, by 2016 the average years of schooling had increased by a year in the country due to various campaigns and probable enlightenment on the need for female education in the region. Following apriori expectation education is likely to increase the probability of farming households adapting to climate changes because it can be assumed that education will increase farm household's awareness of CCA.

The size of households is also shown in Table 1. It is evident that the average household size for the sampled farm households' was approximately six persons in 2010 to eight persons in 2016. Previous literature [35–37] argued that the probability of adopting labourintensive adaptation measures increases with family size due to the availability of free or inexpensive man-power. Also, large families divert part of their labour force into non-farm activities to generate more income [38,39].

From Table 1, the majority of respondents (85.62%) had no access to extension contact in 2010, however, in 2016 the majority (86.79%) had access to extension contact. This showed the intensification of extension services contact increased over the years within the study area. For access to formal credit, the majority (97.5%) of the respondents did not have access; this may limit the ability of the farmers to expand their scale of production. This result is buttressed by the findings of [40] who assessed the trend of formal credit allocation to food crop production in Nigeria; their results showed that there was a decreasing trend in the credit allocation to the food crop production since 2011. It can be argued that agriculture production is constrained in Nigeria by poor credit delivery; the delivery of credit facilities in the country is largely in favour of the wealthy farmers as opposed to poor farmers; the wealthy farmers may utilize the loan acquired for ulterior motives rather than the initial function of agricultural production [40,41]. This limited access to credit facilities may be as a result of high-interest rates credit facilities provided by financial institutions among other bureaucratic delays inherent in loan assessment, acquisition, and disbursement in Nigeria.

Ownership of land can serve as an indicator of the wealth status of farming households [25], and thus it can be expected that the increased wealth status of farming households leads to increased food security of farm households. The distribution of the ownership of land is illustrated in Table 1; the result obtained showed that the majority (98.76%) of farm-households owned land compared to 2010 where the majority (69.96%) of farm households had no access to land.

The summary of average food expenditures and, by extension, food security of farm households, is presented in Table 1. A household is categorized to be very vulnerable to food insecurity if more than 75% of its total expenditure is spent on food items whereas people spending 65–75% are considered to have high food insecurity [11,42]. Following Engel's law, the higher the food expenditure share, the lower the food security of farm households. Thus, the food security status of farm households measured as household food expenditure share slightly reduced in 2016 with average farm households spending approximately 72% of their expenditure on food compared to 2010 when an average farm household spent about 69.6% of their expenditures on food.

#### *2.4. Analytical Techniques*

The method of data analysis adopted in this study to achieve the stated objectives includes:

#### 2.4.1. Panel Probit Model

The pooled probit model specification of the panel data model was employed to evaluate the factors affecting CCA strategies employed by farm households for this study [43,44]. Following Akerele et al. [43], the panel probit model is expressed as follows:

$$y\_{jit}^{\*} = a + X\_{jt} \beta + \varepsilon\_{jit} \tag{1}$$

where

*y*∗ *jit* = the latent (underlying) variable that determines whether farm household *j* would be classified as an adopter of CCA measure *i* at time *t*;

*β* = a vector coefficient;

*Xjt* = a matrix of explanatory variables;

*a* = the constant term; and

*ejit* = the idiosyncratic errors assumed to have zero mean and unit variance. The relationship between the latent variable *y*∗ *jit* and the observed outcome *yjit* is represented as

$$y\_{\text{jit}} = \begin{cases} ^{0 \text{ if } y\_{\text{jit}}^{\*} < 0 \\ ^{\text{if } y\_{\text{jit}}^{\*} > 0 \end{cases} \qquad \text{for } \text{i} = 1, \dots, \dots, \text{...} \text{,} \text{n} \qquad \text{and } \text{j} = 1, \dots, p - 1 \tag{2}$$

where *yjit* = 1 if a farm household adopts a CCA strategy for each adaptation strategy.

The selection of the variables was motivated by previous literature [4,25,45–47], availability of data [48], and economic theories [4] on factors influencing the choice of CCA and are presented in Table 2. Farm and household attributes were included as explanatory variables for assessing determinants of CCA options. Ownership of land or tenancy status was used as a determinant mainly because these can act as proxies for the wealth status of the farm households [25].

**Table 2.** The Description Measurement and A Priori Expectation of the Variables.


Following Teklewold et al. [4], socio-demographic characteristics of farm households important in implementing CCA options were controlled for, these factors include age, gender, household size, gender, and educational level of the household head. Resource constraint was also considered while accounting for the factors influencing CCA options. Quantity of household assets can act as a proxy in measuring the wealth status of farm households [4]; access to credit facilities extension services was included as one of the

explanatory factors since extension services provide crucial education and information needed to adopt CCA options [25,46].

Following from this, the four categories of adaptation strategies were considered in this study, these measures were selected based on the popularity of these measures amongst farm households across all the geo-political zones in Nigeria considered in this study:

Diversify more into other crops Used Irrigation facilities Diversify into off-farm activities Implement soil conservation techniques

Adoption of crop diversification is a CCA method that may involve the planting of high yield variety and drought-resistant crops or intercropping [48], which has been extensively identified from previous studies as an option that can help farmers and farming households buffer the negative effects of climate change. Planting of crops such as cereals that are highly affected with sporadic fluctuations in weather patterns along with turgid crops such as cassava will minimize crop losses due to weather events [33,49].

Implementation of soil conservation techniques such as fallowing and practicing alley cropping can aid in the restoration of soil nutrients, minimize nutrient loss, protects the vegetation cover, and also reduces organic matter oxidation in the soil [49]. Alley cropping can aid in reducing soil erosion while also serving as windbreaks.

Water as a resource is crucial for optimal crops and livestock cultivations [48]. Several studies such as [48–51] documented the importance of implementation of irrigation facilities especially in regions prone to drought or low rainfall occurrences. The adoption of irrigation is encouraged to augment the rainfall amount required for optimal crop cultivation.

Off-farm diversification has been extensively used as a CCA strategy as evidenced by previous literature [51,52]. Farming households may undergo off-farm activities or other occupation during the dry season, unfavourable climate conditions [51] or mainly to complement income sources in order to meet household food security status.

Estimating the impact of climate change adaptation on household food security in Nigeria was achieved in two stages using a combination of two analytical tools, which were described as 2.4.2 (Propensity Score Matching) and 2.4.3 (Difference-In-Difference).

#### 2.4.2. Propensity Score Matching

This study employed PSM to estimate the impact of the adaptation strategies on farm household food security status. The PSM is defined as the conditional probability that a farm household adopts the new adaptation strategies, given pre-adoption characteristics [53]. To mimic a typical randomized controlled experiment, the PSM assumes the unconfoundedness assumption, also known as conditional independence assumption, which implies that once Z is controlled for, technology adoption is random and uncorrelated with the outcome variables. The PSM also accounts for this sample selection bias [25,54].

Following the framework of Ali and Erenstein [25], the PSM was used to estimate the impact of climate change adaptation on farm households' food security. After investigating the choice determinants of CCA practices using PPM, a propensity score matching approach was employed to analyse the impact of adaptation practices on food security.

The farm households were classified as food secure or food insecure based on their share of total household expenditure spent on food. Following Ali and Erenstein [25,42], Smith, et al. [55], households spending more than 75% of their expenditures on food were categorized as food insecure households and were assigned a dummy value of zero; while farm households were categorized as food secure and assigned a value of 1 when the food expenditure is below the threshold level (75%) of total expenditure.

A risk-averse farm Fi opts for a few strategies (*Sj*). It is assumed that households that have opted for adaptation strategies have higher utility levels compared to those that have not: [25].

$$\mathcal{U}[F(\mathcal{S}\_1)] > \mathcal{U}[F(\mathcal{S}\_0)] \tag{3}$$

The PSM can be expressed as according to Ali and Erenstein [25]:

$$\Pr\left(Z\right) = \Pr\left\{I = 1|Z\right\} = E\{1|Z\} \tag{4}$$

where *I* = is the indicator for adoption and

*Z* = the vector of pre-adoption characteristics.

The conditional distribution of *Z*, given *p*(*Z*), is similar in both groups of adopters and non-adopters.

The expected treatment effect for the treated population is of primary significance and is given as

$$
\pi\_{\vert\_{i=1}} = E(\pi \vert I = 1) = E(R\_1 \vert I = 1) - E(R\_0 \vert I = 1) \tag{5}
$$

where *τ* = the average treatment effect for the treated (ATT),

*R*<sup>1</sup> = denotes the value of the outcome for adopters of the adaptation, and

*R*<sup>0</sup> is the value of the same variable for non-adopters.

As noted above, the major problem is that we do not observe *E*(*R*0|*I* = 1), in other words, it is potentially a biased estimator.

After estimating the propensity scores, the average treatment effect for the treated (ATT) can then be estimated as [25,56]

$$\begin{array}{l} \pi = E(R\_1 - R\_0 | I = 1) = E\{E\{R\_1 - R\_0 | I = 1, \ p(Z)\}\} \\ = E\{E\{R\_1 | I = 1, \ p(Z)\} - E\{R\_0 | I = 0, \ p(Z)\}\} \end{array} \tag{6}$$

PSM is based on two underlying assumptions, that is: the common support and the conditional independence assumption [25]. A diagnostic test of matching quality must be carried out after matching to estimate the standard errors and treatment effects. Some balancing tests were to be carried out to access the matching quality, mean absolute bias, t-statics, and the bias reduction before and after matching [57,58].

#### 2.4.3. Difference-in-Difference

DID was also used to assess the impact of CCA on household food security; unlike the PSM which estimates the impact of CCA on household food security between adopters of CCA and non-adopters, DID evaluates the impact of CCA over time, that is from 2010 to 2016.

Difference-in-difference (DID) methods, compared with PSM, assume that unobserved heterogeneity in adoption is present but that such factors are time-invariant. With data on project and control observations before and after the CCA adoption, therefore, this fixed component can be differenced out. Some variants of the DID approach have been introduced to account for potential sources of selection bias. Combining PSM with DID methods can help resolve the problem of selection bias, by matching units in the common support [56]. The propensity score can be used to match participant/adopters and control/non-adopters units in the base year, and the CCA impact is calculated across adopters and matched control units within the common support. For two time periods *t* = {1,2}, the DID estimate for each adoption area *i* will be calculated as

$$DID\_i = \left(Y\_{i2}^T - Y\_{i1}^T\right) - \sum\_{j \in \mathcal{C}} \omega\left(i, j\right) \left(Y\_{i2}^C - Y\_{i1}^C\right) \tag{7}$$

where

*ω*(*i*, *j*) is the weight (using a PSM approach) given to the *j*th control area matched to adoption area *i*.


#### **3. Results and Discussion**

The findings of this research work, interpretations, and also discussion of the result are presented in this section.

#### *3.1. Determinants of Farm Households Climate Change Adaptation Options*

A panel probit model was used in this study to estimate the factors affecting adaptation strategies employed by farming households. Adaptation options identified include

Irrigation Soil conservation Crop diversification Diversification into non-farm activities

The likelihood ratio test from the Panel probit model showed the overall significance of the models at (*p* < 0.01) probability level, which signified that the model is useful in explaining factors influencing decisions of farming households to adapt to climate change.

**Age of Household Head:** As shown in Table 3, the age of the household head is an important determinant in the decision of farming households to use irrigation (*p* < 0.01), and diversify into non-farm activities (*p* < 0.01). The sign of the parameter is negative, implying that the older the household head, the less likely their probability to adopt irrigation and diversify into non-farm activities. It can be deduced from the result that with a year increase in the age of farmers the probability of implementing irrigation facilities and practicing non-farm diversification decreases by 1% and 0.4% respectively. These findings suggest that younger farmers are more likely to adopt these CCA strategies compared to their older counterparts, possibly because they are innovative and keen to try new technology and methods to improve agriculture, whereas older farmers through years of experience may understand the negative economic implications of practicing such strategies. These findings are in support of Ali and Erenstein [25]; where the age of the household head had a negative relationship with CCA adoption, they claimed that older farmers may be conservative about trying new and innovative agricultural practices despite increased awareness. However, the result was against the findings of [59–61] who found that age had a positive association with CCA adoption among farming households.

**Access to Credit Facilities:** Access to credit facilities was positively significant (*p* < 0.01) for practicing soil conservation and off-farm diversification (shown in Table 3), which is in support of Hassan and Nhemachena [62] and Ojo and Baiyegunhi [63]. In their study, they opined variations to farmers' adaptation options, which are largely dependent on their access to credit and information on credit. On the other hand, access to credit facilities has a negative but significant (*p* < 0.10) effect on the probability of using irrigation facilities. This may be due to the cost implication attached to the use of irrigation and the predictive risk of being unable to refund the credit owed when due. The effect on irrigation facilities is in line with [59] but opposed to findings in [64]. Hisali et al. [59] reported that households without credit have a greater likelihood of implementing CCA options, the suggested situations like this may occur where repayment of credit leads to resource constraint needed for CCA adoption or that credit can be used for other purposes other than climate change adaptation.


**Table 3.** Parameter Estimates of Panel Probit Model of Determinants of Farming Households CCA Strategies.

\*, \*\* and \*\*\* represents statistical significance at 10%, 5% and 1% respectively. Authors computation of LSMS data 2010–2016.

**Tenancy Status:** The influence of ownership of farmland is reported in Table 3. As indicated ownership of farmland has a mixed effect on adaptation options, it has a direct and significant relationship with soil conservation (*p* < 0.01) and crop diversification (*p* < 0.01). With ownership of land, the decision on the usage of land rests solely on the farmer, due to the availability of lands, it is easier for the farmer to leave some portion of his land for fallowing and also utilize the farm for the cultivation of crops with varying lifecycles since he does not have to fear he may lose his tenancy status. The cost of incurring land is null, therefore, there are more funds available to go into planting various crops. Quan [65] and Kokoye, et al. [66] concluded that land ownership availability can be an incentive for farmers to invest in resources for farming because farmers can pass their land on to the next generation; therefore, they are more willing to care for the land by adopting practices that can aid to maintain its productivity and food security in the context of climate change. Conversely, ownership of farmland has a negative and significant relationship in practicing alley cropping and diversifying into non-farm activities. The relationship between ownership of farmland and diversifying into nonfarm activities is expected because farmers may have invested so much in their farming business; another reason is that owning land may increase the profitability of the business. Previous studies, however, showed mixed results for the relationship between tenancy status and adoption of CCA options. While some studies [67,68] posit a direct relationship between land ownership and adoption of CCA options, other studies such as [25,69–73] reported a negative correlation. The latter are

variously associated with the need for farmers in this category to have more agriculturally reliant livelihoods.

**Farm size:** Farm size has a significant influence on CCA options. An increase in farm size increases the probability of farmers adopting irrigation (*p* < 0.01), implementing soil conservation techniques (*p* < 0.01), and practicing crop diversification. From Table 3, a 1 hectare increase in farm size increases the likelihood of farm households implementing irrigation, soil conservation techniques, and crop diversification by 2%, 3%, and 3% respectively. Findings are in support of several studies that generally reported a positive association between CCA adoption and farm size [25,70,74,75]. Farmers with large land possessions are likely to have more capacity to try out and invest in climate risk-coping strategies. As reported by Arunrat et al. [64], an increase in farm size and land ownership reduces bureaucratic delays with regards to decisions about CCA adoption, mainly because of their ability to procure the high capital and landholdings, and the freedom required to implement innovative practices on their land.

**Extension Contact:** Studies such as Adams [39], Tambo [51], Boansi, Tambo and Müller [61], and Gbetibouo [76] have shown significant effects of access to extension contact on adopting CCA options. The result of the PPM confirmed that access to the extension has a positive and significant (*p* < 0.01) impact on irrigation use and crop diversification; from Table 3, it can be inferred that a unit increase in farm households' access to extension contact increases the likelihood of implementing irrigation and crop diversification by 17% and 24% respectively. The reason behind it is that extension services help disseminate innovations likely denoting the role of advisory services, and access to information among other resources may motivate the farm household to implement such CCA strategies [70,71]. These findings support those of Tambo [51], Boansi, Tambo and Müller [61], and Gbetibouo [76], which showed that extension services enhanced the availability of information on CCA options.

**Household Size:** It is positive and significant (*p* < 0.01) for the probability of households to diversify into non-farm activities and implement soil conservation techniques. Increasing household size results in an increase in food expenditure and the compulsion to meet this need comes from non-agricultural income sources. Ali and Erenstein [25], Deressa, Hassan and Ringler [45], and Arshad, et al. [77] revealed similar results of the increase in household size, which increases the probability of adopting a strategy. This is likely due to the prevalence of family labour, which makes task achievement more effective, especially during peak periods. Adams [39], Temesgen, Hassan, Tekie, Mahmud and Ringler [47], and Le Dang, et al. [78] contradict the positive influence of household size; they opined that household size has a negative and significant impact on the probability of choosing adaptation strategies.

**Gender of Household Head:** Results obtained in Table 3 are partially in tandem with previous findings [46,50,79,80] that male-headed households often have a higher likelihood of adopting agricultural innovations and thus are better adapted to climate change. Being a male-headed household increases the chances of practicing soil conservation compared to their female counterparts. However, the likelihood to diversify into other occupations increases with being a female-headed household because females in the household especially in Nigeria are found to play supportive roles (such as processors and traders) in the households by diversifying the household income, thus easing the financial burden of the family. Females in households also tend to make financial plans for unforeseen circumstances. Adams [39] and Ogunpaimo, et al. [81] shared a similar view on females tilting towards the adoption of occupation diversification compared to the males.

**Marital status:** Table 3 showed that married farmers have the likelihood to use adaptation strategies such as irrigation facilities, crop diversification, and nonfarm diversification compared to singles. This is likely because more efforts come into making decisions when being married compared to being single. On the other hand, being married has a negative but significant influence on implementing soil conservation techniques.

**Years of schooling:** From Table 3, it is shown that the years of schooling of the farm household head have mixed effects on the choice of CCA. This variable significantly and positively affected practicing diversification into non-farm activities (*p* < 0.01); this result shown in Table 3 supported the work of Ali and Erenstein [25], Alam [48], Gebrehiwot and Van Der Veen [49], and Alam, Alam and Mushtaq [60]. The papers all agreed that educated farmers may be more aware and perceive climate change, as they can easily understand and interpret information compared to farmers with a lower level of education. However, this philosophy did not work for the adoption of some strategies; years of schooling negatively influenced the probability of practicing soil conservation (*p* < 0.01) and crop diversification (*p* < 0.01).

**Quantity of Household Asset:** The quantity of household assets, which is a proxy of the wealth status of farming households, is an important variable that reflects farmers' choice of climate change adaptation options. Results shown in Table 3 contradict Ali and Erenstein [25] in that quantity of household assets enacted a negative influence on climate change adaptation options except for non-farm diversification (*p* < 0.01). It can be implied that income from the use and sale of household size is diverted mainly into non-farm diversification with non-farm diversification serving as secondary income to the farming households.

**Location**: Location typically plays an important role in CCA adoption [25,82–85]. In this study, we included dummies for agroecological zones to control for the location effect on adaptation strategies, with South-South being the base for the model. The result indicated a significant positive and significant probability of farm households in North-Central and North-West to implement irrigation facilities compared to farm households in the South-South region. The likelihood of adopting soil conservation techniques increases with residing in all other regions of the country in relation to the South-South zone. The findings in Table 3 also highlighted that all the zones, except for the South-West zone, negatively affect the probability of farming households to diversify into non-farm activities concerning those in the South-South region.

#### *3.2. Impact of Climate Change Adaptation on Household Food Security*

A combination of PSM and DID was used to evaluate the impact of CCA adoption on household food security. It is therefore imperative to discuss the result of the impact of CCA options on household food security between adopters and non-adopters from 2010 to 2016.

#### 3.2.1. PSM Result of Impact of CCA on Household Food Security

The with and without effect of climate change adaptation options is explained by PSM. Table 4 presents the impacts of adaptation methods used on household food security based on propensity score matching. The impact of climate change adaptation on household food security was significant with adopters having 9% higher food security than non-adopters in 2010. This result is in support of Ali and Erenstein [25], but against Weldegebriel and Prowse [86] who found that the adaptation strategy reduced farm income and, with that, food security due to the exclusion of important variables. Ali and Erenstein [25] stated that CCA practices help to enhance the food security and welfare of rural households. Thus, farm households should be encouraged to adopt a few CCA practices to improve welfare outcomes. Farm households not adopting CCA practices are more likely to be food insecure. Shiferaw, et al. [87] also opined that the average treatment effect on the treated (ATT) of adaptation on household food security was positive and significant, which implied that CCA options foster household food security.


**Table 4.** PSM Showing the Impact of CCA Adoption on Household Food Security.

Authors computation of LSMS data 2010–2016. \*\* represents statistical significance at 5% respectively.

Table 5 shows the covariate balancing tests before and after matching. As indicated in the table, the balancing test revealed that the bias was relatively higher before matching.


Authors computation of LSMS data 2010–2016.

For instance, before matching tenancy status (*p* < 0.10), extension contact (*p* < 0.05) and gender of the household head (*p* < 0.01) could cause selection bias when assessing the influence of CCA options on household food security status. The percentage bias reduction is between 65–97.2%. These indicators of covariates balancing showed the results obtained satisfied the balancing of covariates following matching and the application of the common support condition. The result implied no selection bias when matching adopters and non-adopters, thus differences in food security levels are mainly due to the adoption of CCA measures.

#### 3.2.2. DID Result of Impact of CCA on Household Food Security

The true impact of CCA on household food security over time can be measured by looking at the effects of adaptation options between adopters and non-adopters, which was illustrated by the result of the PSM and then measuring the impact of the adaptation measures over the period of adoption using the DID. The adopter and the non-adopter groups within the same common support in the PSM analysis for the base period were appended, after which the DID analysis was carried out.

The difference in difference (DID) estimation combined with propensity score matching (PSM) was used to evaluate the average impact of the CCA options on household food security. The average treatment effects of CCA options were evaluated, which compares food security in the adoption state (Y1) with the outcomes in the control or the counterfactual (Y0) conditional on receiving treatment.

Contrary to previous studies [25,88], which used cross-sectional data to assess the impact of CCA options, this study used panel data for six time periods to assess the impact of CCA options on household food security. Similar to Kangmennaang et al. [26] and Kabunga, et al. [89], this approach allowed for the combination of propensity score matching with DID estimation to control for selection bias and temporal impact variability. The estimated results showed that adopting CCA options intervention positively influenced household food security.

As shown in Table 6, the F-value is significant at (*p* < 0.01), which indicated that the model was useful in assessing the impact of CCA on household food security over time. The result in Table 6 showed that the coefficient of time trend (y16) was significant (β= −4.02, *p* < 0.01); this implied that household food security was trending down with time. The result of the DID is positive and significant (β = 5.93, *p* < 0.01), which reveals that the impact of the CCA options increases household food security between adopters and non-adopters. This finding confirms that CCA options had a significant positive impact on farm households' food security status. This finding shared similar results with Noltze, et al. [90], and Kangmennaang et al. [26] who found that agroecological practices in the form of CCA promote food security after 2 years. However, while this study adopted the use of HFES as a measure of food security, Kangmennaang et al. [26] used the Household Food Insecurity Access Scale (HFIAS).


**Table 6.** DID Showing the Impact of CCA Options on Household Food Security Without Covariates.

Authors computation of LSMS data 2010–2016. \*\*\* represent statistical significance at 1% respectively.

It must be noted that CCA adoption may be implemented by farming households before the year 2010, however, 2010 was used as the baseline due to data availability. The findings in this study also support the result of other studies that confirmed the direct effects of CCA options on household food security. Becerril and Abdulai [91] reported that increased farm output can lead to higher consumption, off-farm diversification, and increased farm incomes. Surpluses from farm yield may also be used to increase the household quantity of assets increasing the adaptive capacity of households to climate change, thus promoting households' food security status [92,93]. Khonje, et al. [94] also reported that sustainable practices such as crop diversification and other CCA options can lead to improved welfare and food security outcomes. Other effects of adopting CCA options reported may include promoting women empowerment, capacity-building, and knowledge exchange within the community, which may further lead to increased food production at the community level, increased consumption, and better living standard conditions [26]. Adopting CCA options can also foster collective relationships among farming households within the communities, allowing for the sharing of risks and burdens associated with farm activities.

To control for any selection bias between the adopters and non-adopters of CCA, the results obtained in Table 6 were controlled for covariates influences on the impact of CCA on household food security status, as shown in Table 7. The result obtained indicated that even after controlling for potential covariate influence, DID had a positive and significant (β = 4.15, *p* < 0.01) effect on household food security status. This result was corroborated by [26] who found that covariates control does not influence the outcome of the DID result or the influence of CCA on the household food security status.


**Table 7.** DID Showing the Impact of Adaptation Options on Household Food Security with Covariates.

Authors computation of LSMS data 2010–2016. \*\*\* represent statistical significance at 1% respectively.

#### **4. Conclusions**

This study assessed the impact of climate change adaptation (CCA) on household food security among farm households in Nigeria. Against previous works of literature that adopted cross-sectional approaches to investigate CCA impacts on welfare outcomes, this research work adopted a panel data analysis, thus measuring the impacts of CCA on household food security across space and time. We recognized that there are other CCA options not considered in this study mainly due to lack or limited data of such CCA strategies in the LSMS data. However, this study has provided useful insights and information on the relationship between CCA options and household food security in Nigeria. Based on the aforementioned findings, this study confirmed the need for adaptation to climate change by farming households, which increase with an increase in farm size, extension contact, and marital status, with access to credit, ownership of farmland, household size, the gender of household size, years of schooling, household asset and location having mixed effects on the choice of adaptation strategies. From the study, it was shown that climate change adaptations have helped farming households improve their food security status in the face of prevalent climatic conditions. Therefore, the study recommends that farming households should practice continual implementation of CCA options to foster improvement in household food security status. Also, access to credit facilities and extension contacts remains a catalyst for implementing adaptation measures, thus constant and quality extension contacts and credit facilities with low-interest rates should be provided to farming households to enable them to adapt better to climate changes and improve household food security status.

**Author Contributions:** Conceptualization, O.R.O.; Formal analysis, O.R.O.; Methodology, O.R.O.; Software, O.R.O.; Validation, Z.O.-U., and J.S.; Writing—original draft, O.R.O.; Writing—review & editing, Z.O.-U., and J.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Publicly available datasets were analyzed in this study. This data can be found here: [https://microdata.worldbank.org/index.php/catlog/lsms/] at reference number [Ref: NGA\_2010\_GHSP-W1\_v03\_M, Ref: NGA\_2012\_GHSP-W2\_v02\_M, Ref: NGA\_2015\_GHSP-W3\_v02\_M].

**Conflicts of Interest:** No conflict of interest was declared by the authors of this paper.

#### **References**


### *Article* **Mainstreaming Disaster Risk Reduction into Local Development Plans for Rural Tropical Africa: A Systematic Assessment**

#### **Maurizio Tiepolo \* and Sarah Braccio**

Interuniversity Department of Regional and Urban Studies and Planning (DIST)-Politecnico and University of Turin, viale Mattioli 39, 10125 Turin, Italy; sarah.braccio@polito.it

**\*** Correspondence: maurizio.tiepolo@polito.it; Tel.: +39-0110907491

Received: 21 January 2020; Accepted: 9 March 2020; Published: 12 March 2020

**Abstract:** Disaster risk reduction in rural Africa can contribute to reducing poverty and food insecurity if included in local development plans (LDPs). Five years after the Sendai Framework for Disaster Risk Reduction (DRR), we do not know how much risk reduction is practiced in rural Africa. The aim of this assessment is to ascertain the state of mainstreaming DRR in development planning in the rural jurisdictions of tropical Africa. One hundred and ninety-four plans of 21 countries are considered. Ten characteristics of the plans are examined: Climate trends, hydro-climatic hazards, vulnerability and risk assessments, alignment with Sendai Framework, vision, strategies and objectives, DRR actions, internal consistency, DRR relevance and funding sources, local and technical knowledge integration, public participation. It is found that local climatic characterization is almost always absent and risk reduction is an objective of the plans in one case out of three. Prevention actions prevail over those of preparedness. There is poor participation in the plan preparation process and this limits the implementation of the actions. A modification of the national guidelines on the preparation of LDPs, the orientation of official development assistance towards supporting climate services and the training of local planners, together with the increase of financial resources in local jurisdictions are essential for improving DRR at local scale.

**Keywords:** disaster risk reduction; official development assistance; public participation; risk tracking; rural development; Sendai framework; sustainable development

#### **1. Introduction**

Tropical Africa presents specific characteristics as opposed to other regions of the Global South. Firstly, 62.5% of the population is still rural [1]. Over half of jobs and 69% of income come from agriculture [2]. Despite this, 58% of the population is in conditions of food insecurity [3]. Poverty and inequalities between countries and within individual countries reach the world's highest levels [4]. In these conditions, the development of agriculture is considered better for absorbing the poor than industry and services [5–9]. However, agriculture is strongly exposed to climate change (CC), which affects rain-fed crops and livestock [10–12], casts smallholders into deeper poverty [13], and makes investments vulnerable [14] because of low adaptive capacity [15]. Finally, fluvial flooding and dust storms affecting urban areas [16–18], are formed in the surrounding rural areas and it is there that they should be treated primarily [19]. For all these reasons, rural environment remains the hot spot for disaster risk reduction (DRR) in Africa.

Smallholder farmers from tropical Africa have insufficient financial capital and willingness to change to significantly reduce disaster risk [20] and sometimes lack natural, social, and human capital [21]. DRR should first be addressed by the local governments of all those countries where the process of administrative decentralization has assigned them the task of conserving the environment and protecting the population from natural hazards. In addition, precisely for this reason Official Development Assistance (ODA) helps local governments strengthen their capacities with specific programs [15]. The substantial reduction of disaster loss and damage and the increase of local disaster risk reduction strategies by 2030 has become a target of the Sendai Framework for Disaster Risk Reduction (2015) [22,23]. Five years after the Sendai Framework we only know the number of African countries with DRR strategies in place: Just seven, according to UNDRR, which is in charge of monitoring the implementation of the Sendai Framework [24]. We do not know if this results from a lack of information or from national strategies which "often do[es] not penetrate to the local level" [25].

However, the number of strategies in place does not guarantee a reduction of risk at local scale. Peer-reviewed literature points out that DRR mainstreaming into local development plans (LDPs) is often based upon rough climatic analyses, produced with little public participation, done with few actions, mostly of little significance, and lacks sufficient resources for implementation [26–39]. Genuine public participation in the planning process is critical to reducing disaster risk. This is particularly true in rural areas, where almost all actions require the direct involvement of smallholder farmers and herders [40–45].

The problem is, therefore, to know the content of the DRR strategies in place locally and the process by which they were prepared, rather than just knowing their number. An evaluation of the local DRR strategies can therefore highlight the aspects to be improved in the DRR and inspire the ODA consequently.

The aim of this article is to ascertain the mainstreaming of DRR in local development plans for rural areas in tropical Africa.

We will proceed by identifying the local jurisdictions. To compare the LDPs, it is necessary to investigate a homogeneous climatic area. The tropical zone is the most present in Africa. Within it, the analysis is restricted to English-speaking, French-speaking, and Portuguese-speaking jurisdictions with plans in force: Municipalities, cantons, districts, counties, sometimes regions. The LDPs are identified by Google search. The plans considered are only those in force at September 2019 in jurisdictions with at least 90% of the territory in the tropical zone and rural population exceeding 50%. With these restrictions, we obtained 194 plans, covering 24% of rural jurisdictions in 21 tropical African countries.

Ten characteristics of the plans are considered: (i) Climatic trend, (ii) hazard identification, (iii) existence of vulnerability and risk assessments, (iv) alignment with the Sendai framework for DRR, (v) vision, strategies and objectives, (vi) DRR actions, (vii) internal coherence, (viii) budget size and origin, (ix) local and scientific knowledge integration, (x) public participation in the plan preparation.

Although DRR mainstreaming in the LDPs is in place in half of the countries of tropical Africa, improvements are needed in analysis and planning, as well as resources to achieve the actions, if DRR is to be consolidated at local scale.

The importance of this article is in highlighting the weaknesses of today's local DRR. However, also in proposing a simple assessment framework, repeatable over time and on a larger scale, which was still missing. These features can facilitate the transition from occasional assessments to tracking the DRR in local development plans.

#### **2. Materials and Methods**

The assessment is split into four phases. The first phase identifies the jurisdictions being studied. We begin by outlining the tropical zone according to the Köppen-Geiger classification (tropical rain forest, tropical monsoon, tropical wet and dry or savanna, hot and hot-semi-arid desert climates) based upon temperatures and precipitation, as calculated in the period 1980–2016 on a one kilometer grid [46], which updates, in greater spatial resolution and over a more recent timeframe, the work of Rubel and Kottek [47]. We then identify the government level responsible for producing local development plans. In some countries, it is the municipality, in others it is the district, the canton, the county, and

sometimes the region. In Africa, these different jurisdictions do not necessarily correspond to territories of growing extension: A municipality in Cameroon, for example, may be larger and more populated than a district in Uganda. The jurisdictions of Angola, Ethiopia, Kenya, Madagascar, Malawi, Namibia, South Africa, Tanzania, Uganda, Zambia and Zimbabwe, which fall only in a small part in the tropical zone, were excluded. Finally, we considered only the jurisdictions with rural population exceeding 50% (Figure 1).

**Figure 1.** Flow chart of the assessment.

During the second phase, the plans are obtained. We searched Google with the name of each local jurisdiction and the integrated, local, municipal or strategic development plan and obtained 194 plans of 21 countries of which 107 were prepared after the Sendai framework for DRR, therefore likely to integrate its principles (Figure 2).

We also obtained the national strategies and guidelines for the preparation of the LDPs for each of the 21 countries.

The third phase identifies the information required to examine the 10 key characteristics of the plans: Climatic trend, hydro-climatic hazards, vulnerability and risk assessment, Sendai alignment, vision, strategy and objectives, actions, internal consistency, DRR budget and source, local and scientific knowledge integration, public participation (Table S1).

During the fourth phase, the information is collected, processed, and analyzed. The analysis of the climatic characterization ascertains if the plans are based on a series of at least thirty years of climatic observations and trace long-term local climatic scenarios. It is ascertained whether the plan is based on a CC vulnerability or risk assessment of the local communities. The alignment of the LDPs with the Sendai Framework verifies that the plan is subsequent to the National DRR Strategy and checks that at least the DRR and Sendai terms appear among the words of the plan. The analysis of the vision, strategy, and objectives verifies if DRR is present. The actions are broken down by purpose: Prevention, namely "to avoid existing and new disaster risks", and preparedness, i.e. "to recover from the impacts of likely, imminent or current disasters" [48]. Therefore, the actions are compared with those recurring in the literature on the individual hazards [49–60] to verify their completeness. The internal consistency analysis ascertains whether the plan relates to the climate trend, impacts, and possible solutions using specific frameworks. Secondly, the actions proposed by the communities are considered to verify if they subsequently appear among those retained by the plan and prioritized for the first year. The analysis of the budget reserved to DRR actions follows. In this regard, it is important to ascertain the share of the budget reserved for the DRR and to what extent it comes from the local administration's own resources, from the central government or from donors. The planning process is then examined to discover if local and technical knowledge has been integrated. Finally, the role of public participation in preparing 105 municipal development plans is ascertained: Analysis, planning, adopting, implementing, monitoring and evaluating (M&E) the plan while representing

the many communities that make up a rural jurisdiction and gender constitute as many levels of participation [41].

**Figure 2.** The 194 local jurisdictions (LG) in the tropical zone (T) with development plans in force and subtropical (ST) and boreal (B) zones.

#### **3. Results**

Monitoring the implementation of the Sendai Framework offers little information on the rural jurisdictions of tropical Africa. Considering DRR mainstreaming in 194 LDPs in force in half of the African countries gives a vision of 1.2 million km2 populated by 37 million inhabitants. The rural jurisdictions considered, irrespective of the level (municipality, district or canton, county, region) are vast. The municipality of Kagisano Molopo (South Africa) for example, spreads across 23,800 km2, that of Mintom (Cameroon) over 11,000 km2. Jurisdictions contain an average of 52 settlements, as well as the municipal capital town and an average population of 193,000 inhabitants. Vast territories present several micro-climates and hazards that differ from one area to another, to be addressed with specific actions. In addition, the presence of many communities makes the representation during the plan preparation process a challenge.

#### *3.1. Local Planning Overview*

The local jurisdictions use different plans depending on the country: Municipal (district or regional) development plan, municipal (or county) integrated development plan, municipal (district or regional) strategic plan. Irrespective of the name, the plans have some common characteristics. Firstly, they are medium-term tools (3–5 years) prepared following national guidelines. Secondly, the plans

are split into two parts: A factual base (with climate section) identifies the problems and available resources, then the actual planning, which usually defines the vision, strategies, objectives and actions. The latter are localized, quantified, their cost is defined, as well as the origin of the funds to be used to finance them. Some LDPs take stock of the previous plan. Other plans present the actions necessary for each community and the priority actions to be implemented during the first year. The strategies never detail the actions (quantity, costs). In the five countries with the highest number of plans, those in force and freely accessible on the web cover from 12% of the rural districts of Uganda to 94% of rural municipalities of South Africa (Table 1, Figure 3).


**Table 1.** Local development plans (LDPs) considered in five countries.

The Sendai Framework schedules the alignment of local DRR strategies with national ones. In fact, some countries have a national strategy (Cape Verde, Madagascar), others have a national plan or a preparation, management or response policy to the risk (Burkina Faso, Kenya, Mozambique, Namibia, South Africa, Uganda) or a national plan to strengthen DRR capacities (Chad, Niger). These heterogeneous tools at national scale define some important elements for local planning: The legal framework, the competencies of the different players, the operational procedures, the coordination between players in the case of emergency, the information path in the case of early warning, the capacities to be strengthened, the awareness-raising and participation, the methods of post-disaster recovery, and the funds to implement all of this. Today, out of the 21 investigated countries, 10 do not have a national strategy, three countries (Benin, Cameroon, and Mozambique) have LDPs already in force before the national strategy and only eight countries have a strategy following which a new generation of LDPs was formulated, thus being in line with the Sendai Framework (Figure 4). Without a national strategy, the local plans would be deprived of the aforementioned elements.

**Figure 3.** Cameroon (1), Ghana (3), Kenya (2), South Africa (5), Uganda (4). Ongoing local development plans in tropical zone (P) at September 2019.

**Figure 4.** Ongoing LDPs (black segments), national disaster risk reduction (DRR) strategies (red dots), and the five years' time to attend targets of the Sendai framework for DRR.

#### *3.2. Analysis Phase*

#### 3.2.1. Climate Characterization and Hazards

LDPs rarely characterize the climate as they consider too short a timeframe, or they are based only on local knowledge. In Niger, the latest generation LDPs are the exception: Precipitation and temperatures analyzed over a thirty-year series of data, identification of dry spells, sometimes of wind. Increase of temperatures, heavy rainfall, late onset of the wet season, and decline in total annual rainfall are the changes most frequently observed (Table 2). The CC local scenarios are never presented unless the national ones are reported. Conversely, the LDPs almost always identify the main hydro-climatic threats. Drought is by far the hazard most frequently reported by the plans. This is followed, with less frequency, by strong winds, bush fires, and floods. High temperatures, heavy storms, sea level rise, lightening, hailstorm, dust storms, salt-water intrusion and cyclones are reported more rarely (Table 3).

**Table 2.** Main local climate trends referred by 84 LDPs in rural Africa.



**Table 3.** Main hazards referred by 176 local development plans in rural Africa.

#### 3.2.2. Vulnerability and Risk Assessments

The considered LDPs do not have vulnerability assessments (quantification and localization of the vulnerable population to the individual hazards) and the very term "vulnerability" is found only in 43% of the plans. Even the risk assessment (probability of occurrence of the individual hazards, risk level, risk mapping) is absent. Moreover, the term hazard appears in only 46% of the plans.

#### *3.3. Planning Phase*

The LDPs mention the Sendai framework in 40% of cases and DRR in 46% of cases, although they include some typical DRR action without being defined as such. Many plans provide a long-term vision. The most frequent envisage a prosperous, wealthy community, which has achieved sustainable development and a high quality of life. The strategies proposed to achieve this vision involve the construction of roads, water, sanitation and hygiene (WASH), afforestation and use of green energy. The most frequent objectives remain the disaster prevention and response, the construction and maintenance of roads, increased access to drinking water, to basic services, to electricity (Table 4).


**Table 4.** Most common vision, strategies, and objectives in LDPs for rural Africa.

As to the actions, those of DRR prevention (119 plans) prevail over those of preparedness (49 plans). The most frequently illustrated actions of hydro-climatic DRR are afforestation (50%), the use of drought tolerant crops (23%), the construction or rehabilitation of boreholes and wells (22%) and CC sensitization (18%). The preparedness actions are much less frequent: Disaster relief (24%), early warning (24%), and disaster management plan (18%) (Table 5).


**Table 5.** Frequency of risk prevention and preparedness actions in 127 LDPs in tropical Africa.

We note the absence of some canonical actions for addressing the individual hazards. With respect to drought, there is no weather forecasting actions to encourage run-off infiltration, crop residues stocking, micro-credit, and self-help community groups [52], no crop insurance [60]. In relation to strong winds, there is no house retrofitting. There are no actions to deal with bush fires with respect to the many already trialed elsewhere in prevention (vulnerable population localization, alert protocols, access routes, lessening fuel pressure in forest areas) and in preparedness (fire break around homes) [53,54]. Half of the countries considered overlook the Ocean. Sea level rise threatens long stretches of West African coast [61]. The plans do not envisage any measure to reduce the risk of coastal flooding, such as natural buffers, beach nourishment, sea walls, and elevated houses [51]. Apart from the reduction of emissions, which has a long-term effect on the local climate, and afforestation, there are no other actions aimed at reducing the impact of heat waves. There is no early warning, no heat wave action plan. The prevention of dust storms with the stabilization of dunes with straw checkerboard and the increase of arboreal and herbaceous vegetation on denuded soils, for example, using wheatgrass, are not planned [55–57]. There are no measures to address salt-water intrusion consequent to sea level rise [58,59]: A frequent phenomenon in coastal areas that affects agriculture and access to drinking water. As to preparedness in general, there are no climate scenarios. Civil protection is mentioned by just 21% of the plans. There is no mention of simulations and drills, and even less so, actions informing the public on what to do in case of a warning (Figure 5). Resilience is a frequent word of the plan (65%).

The LDPs use different instruments to control the coherence between problems identified, objectives, and actions. The most complete are proposed by Cameroon's LDPs with the climatic threat-impact-strategies-actions framework and the logical framework (objective-outcome-indicator) (Table 6). Despite this, the DRR actions proposed by the individual communities rarely appear among the priority ones of the plan (Figure 6).

**Figure 5.** Fifty-seven actions explicitly addressed to DRR in 111 LDPs for rural tropical Africa (figures stand for action frequency) and missing actions (red).

**Table 6.** Tools for threat-actions consistency control in 100 LDPs of tropical rural Africa.


**Figure 6.** Consistency of climate change (CC) actions between the planning phases in Cameroun's LPDs.

The resources scheduled by the plans in the medium-term to implement DRR vary greatly from country to country, but in the majority of cases they are below 20% of overall expenditure. Benin, Niger, and South Africa have plans that reserve the highest share for DRR. Rwanda, Djibouti, and Malawi allocate the lowest share (Figure 7). Local governments never finance over 10% of DRR actions. The remaining 90% is borne by the State or by donors.

**Figure 7.** Budget (%) for hydro-climatic DRR in 13 African countries.

#### *3.4. Plan Preparation Process*

Three aspects of the plan preparation process are relevant for actions implementation: The integration of scientific knowledge with local knowledge, public involvement, and gender representation.

The main fields in which scientific knowledge can contribute to LDPs are climate characterization, hazard probability of occurrence, hazard prone zones identification, risk level and assessment within a back-casting exercise (expected effects of actions), definition of some preparedness actions (e.g., early warning system). In none of these fields, with the sole exception of Niger's LPDs, do the plans use scientific knowledge: The information is extracted exclusively from local knowledge using participatory rural appraisal tools in the analysis phase.

The information contained in the plans allow for two aspects of participation to be appreciated. First, the community, community-based organizations (CBOs), and individual citizens participation in analysis, planning, implementation, plan monitoring and evaluation (M&E). Second, gender

representation in the analysis, plan approval, and M&E. The analysis is developed only on the municipal development plans, which are the real arena on which participation can develop.

Local jurisdictions in rural Africa contain many communities. On average, the municipalities have 52 human settlements. As a consequence, in many cases, the participatory analysis process occurs by bringing together the delegates of each community into zone centers. In the cases where the plans provide information on this phase of the preparation process, the participation attends 81% of the plans considered. In this phase, the communities provide information on their needs, sometimes solutions, expressed in the best of cases as priority micro-projects (or actions). Public participation of individual communities and individuals is explicitly required by 70% of the plans to implement afforestation and health actions (vaccinations, construction of health centers) and in financing of actions mainly. However, only one out of three plans involve communities in strategies, priority actions identification, and budgeting. Planning is reserved to the municipal council, sometimes expanded to ministerial representatives and economic operators. Representatives of CBOs and communities are excluded from the M&E committee in two plans out of three (Table 7).


**Table 7.** Public participation in municipal development plans preparation in rural tropical Africa.

Almost all of the plans call for greater participation of women and sometimes of young people and minorities in the decision-making processes. However, there is little understanding of how to achieve it. Furthermore, awareness-raising is proposed, rather than organizing activities at times that allow women to participate and to lighten the workload on their shoulders, increase the level of education. In fact, the plans contain little information on how gender involvement occurred in the preparation process. In the analysis phase, the share of women in community delegations is on average 25%. The adoption of the plans is the responsibility of the city council. In this assembly, gender representation is on average just 28%. However, the differences in gender representation from country to country are large: Senegal (44%), South Africa (43%), Cameroon (24%), Niger (16%), Burkina Faso (12%), Benin (5%) (Table S1). Monitoring and evaluation of the plan is the responsibility of the M&E committee, a body in which gender representation drops to 16% (Table 8).

**Table 8.** Gender participation in the municipal development plans (MDPs) preparation in rural tropical Africa.


#### **4. Discussion**

In Africa, DRR is particularly important for protecting the primary sector from the impact of climate change and thus to allow agriculture, breeding, and forests to reduce poverty and food insecurity, which remains the characteristic trait of the Continent compared to other regions of the Global South. Unfortunately, the monitoring of the Sendai Framework carried out by UNDRR so far tells us little about the state of DRR in rural Africa. Through our systematic assessment, we have ascertained that almost all countries have LDPs at the scale of the municipality, district/county, canton, or region. In half of the countries, those instruments are freely accessible on the web. The analysis of 10 characteristics of the plans has allowed us to characterize the local DRR.

The plans not having climate analyses do not identify the hazards according to the probability of occurrence, and they do not outline climate scenarios. These deficiencies are serious when considering the vastness of the jurisdictions considered, which require a spatial characterization of the climate. Exceptions aside, hydro-climatic threats are identified based solely upon local knowledge and are not positioned hierarchically by severity or frequency. Vulnerability and risk assessments are lacking.

The planning process does not integrate scientific and local knowledge and presents problems already observed in other contexts. The plans prepared internally to the municipality are of higher quality than those prepared by consultants [62].

The preparation process of the LDPs at a municipal scale begins with the identification of the needs of the individual communities. When this occurs in territorial assemblies attended by the delegates of the communities of the zone, the representation of the communities and of gender is low. This is the only arena in which public participation takes place and confirms what has already been observed ten years ago with respect to the process of preparing the first generation local development plans in the Sahel [63].

Only in a third of cases do communities decide on the municipal development plan and participate in monitoring and evaluation activities. In the rest of the cases the delegates of the community do not appear in any committee, do not participate in any negotiation, do not receive any delegation of power (empowerment) [64], and do not control anything, not even the stage of progress of the plan [41].

The true planning occurs in another arena, which is accessed by the municipal councilors, technical services, representatives of the ministries, and sometimes donors. Here, gender representation is 28% only. Planning usually begins with a visioning exercise which is not, however, followed by that of back-casting: Estimating how many actions would be necessary to significantly reduce the climate risk and proceeding backwards to identify how many of them to carry out with the plan. This exercise is impeded by the lack of knowledge on the frequency of occurrence of the hazards, on the exposed zones, on the impact of DRR actions. For example, in relation to pluvial floods, we do not know how much the run-off reduces on different types of soil by virtue of infiltration works (trapezoidal bund, half-moons, stone lines, etc.). Therefore, it is impossible to estimate the reduction of risk following the risk treatment. The objectives are dominated by efforts to reduce the hydro-climatic risks, then to break isolation (roads) and meet primary needs (WASH, electricity).

When risk prevention actions are planned, the most frequent concern afforestation (as a means of conserving soil), agriculture and access to water (drought prevention). Different plans (Kenya) contain mitigation actions (renewable energy, energy saving/LED, improved stoves). However, mitigation actions are not able to contain the rise in temperatures, for which land-based mitigation (vegetation) and interventions on building materials are required [29]. Ultimately, the priority actions contain little DRR: The exposed zones are not identified, the dynamics of the settlements within them are not known, early warning systems are infrequent or not designed involving populations at risk [65], house retrofitting is not facilitated. School education as a public awareness channel [28] is not practiced, apart from tree planting.

The planned climate actions are never financed by the municipalities over 10%. The remainder is borne by donors or is simply not funded. The financial weakness of rural local administrations [33,36] is also confirmed in tropical Africa. In these conditions the implementation of the plan falls largely on the shoulders of communities and individuals, which are asked to finance the actions or to provide materials and labor but in two thirds of cases they have no voice in the planning process. Although the plan preparation process lasts an average of 10 months, genuine public participation is infrequent.

The literature is filled with examples in which the lack of participation translates into a lack of implementation [36,38,41]. Several local governments are planning to activate a municipal website. However, uploading the LDP on the web remains the most common E-government action [66,67]. Simple information systems through smartphones are not used, with the sole exception of Senegal.

The monitoring indicators used by UNDRR do not measure the local capacity of DRR. It is not sufficient to know country by country the percentage of local jurisdictions with DRR strategies aligned with the national DRR strategy. It is necessary to know the content of those strategies, how they are formulated, what potential they have to be implemented. Widespread poverty [9] and food insecurity [3] lead many local governments to support primarily agro-pastoral production and to meet primary needs which are still unsatisfied (WASH, electricity, health, education), before dealing with DRR. By doing so, the primary sector remains exposed to hydro-climatic hazards.

The assessment has produced four unexpected findings. Firstly, the large number of LDPs in force (and more are still not freely accessible). Secondly, a very articulated palette of hazards, despite being dominated by the ubiquitous drought. Thirdly, the emergence of some actions still not widespread but important: Risk reduction local funds, considered important instruments for reducing poverty [9,13] and the use of renewable energy sources. The latter have the benefit of being present in situ (sun, wind, water), do not require transportation costs and are not subject to price increases (petrol), factors that penalize the smallholder farmers of remote rural areas. Fourthly, the lack of genuine public participation, despite this term being in many plans among the most frequently used.

The main limitation of the assessment is the number of plans examined, moreover relating to just half of the Continent's tropical countries, of which only 107 plans were formulated after the Sendai Framework for DRR. Another limitation is the fact of considering only planning and not implementation.

The major problems that a multicountry and multilanguage assessment has to face is the understanding and comparison of the plan budgets, which sometimes provide little detail (actions merged by sector, absence of total cost, etc.) and of the participation process.

The implications of this assessment concern DRR mainstreaming in the next generation of LDPs. We have observed some components of the LDPs critical for DRR. With regards to the use of the assessment results, three recommendations apply.

The first recommendation concerns the national guidelines for the preparation of LDPs. If DRR mainstreaming is to be improved in the next generation of plans, the guidelines should require climate characterization and mobilize national meteorological services to provide climate services to local governments. The plans should identify the zones (and inhabitants) exposed to hazards with greater probability of occurrence, quantify DRR, and estimate how far the risk is reduced if the actions scheduled in the plan are implemented in the 3/5 years of planning, introducing the back-casting exercise. The quantity of tables required should be reduced in favor of those strictly necessary to control the coherence between threats and priority actions. Plans that do not include DRR hydro-climatic actions among the priority ones must motivate this decision. Finally, the guidelines should require a precise description of the plan methodology and the representativeness of communities and gender in all phases of the process: Analysis, planning, adoption, and M&E.

The second recommendation concerns the ministries of those countries that have still not made the LDPs freely accessible on the internet. The portals from which it is possible to access the plans of Cameroon, Ghana, Kenya, Namibia, and South Africa are best practices to be used for inspiration.

The third recommendation concerns ODA. Donors should consider supporting the revision of the national guidelines for LDPs preparation, climate services to local administrations (including the upgrading of the local weather stations and early warning systems), and strengthening the capacity of local planning units.

These three recommendations, if implemented, would improve DRR at local scale. Monitoring the number of local risk reduction strategies does not help to understand if a substantial reduction of losses and damage and increase of disaster risk strategies is being achieved.

#### **5. Conclusions**

Five years after the Sendai Framework, we do not know whether rural municipalities in tropical Africa have truly succeeded in mainstreaming DRR in their LDPs. The problem is knowing the quality of these plans, rather than the number of local strategies in place. We, therefore, considered the mainstreaming of the DRR in 194 LDPs of rural Africa post-Sendai, observing ten characteristics of these tools.

With a few exceptions aside, we noted the absence of climate characterization and scenarios. Vision, strategies, and objectives rarely mention DRR. LDPs strengthen livelihoods (agriculture, livestock, fisheries), increase the access to basic services (WASH, electricity) and to infrastructures to break the isolation of remote communities. Some actions in these sectors are also of DRR (access to drinking water and watering) but are not enough to face sea level rise, bush fires, dust storms, salt- water intrusion. Civil protection and crop insurance are missing from preparedness actions. The DRR actions identified by the communities rarely become priority actions of the plan. Public participation is more a goal than the approach followed to prepare the plan. The plan requires the participation of the communities and individuals to be implemented but the national rules and sometimes local administrators exclude genuine public participation in the planning process which are consequently poorly representative of the communities and gender. It should be remembered that gender share among the municipal councilors stops on average at 28%. Only Senegal and South Africa approach the correct gender representation of municipal councilors. This inevitably limits the plan implementation. The planning process, therefore, needs to be reviewed if DRR is to be fully integrated at local scale and translated into solid appropriate actions by the local communities.

In developing the assessment, we considered plans that sometimes provide little information about the preparation process and budget. This restricted the sample considered to analyze these characteristics, limiting the significance of the relative results.

In many tropical African countries, it is not possible to increase significantly DRR by the action of individual smallholder farmers and herders. For this reason, DRR is an institutional task of local governments, since the first steps of the administrative decentralization process started twenty years ago in many countries of tropical Africa. The main tool used for this purpose is the LDP, as it is a medium-term planning tool, mandatory by law and with a long tradition. However, the LDP has weaknesses that should be identified to be reduced. If this is not done, disaster risk reduction will not take root and CC and variability will continue to generate food insecurity and poverty.

The problem of not knowing the content and process of the DRR strategies in tropical Africa has been addressed by considering 194 LDPs using an assessment framework that can be repeated in other countries. We recommend that any organization willing to support local disaster risk reduction should consider the assessment framework proposed in this study and use it to switch from an occasional assessment to a tracking process. It will thereby be possible to continue to have an understanding of improvements and residual fragilities in local DRR rather than merely counting how many strategies are in force.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2071-1050/12/6/2196/s1, Table S1. Key characters of 194 local development plans for rural tropical Africa freely accessible on the web.

**Author Contributions:** Conceptualization, M.T.; methodology, M.T.; investigation, M.T. and S.B.; writing—original draft preparation, M.T.; writing—review and editing, M.T. and S.B.; visualization, S.B.; funding acquisition, M.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by DIST-Politecnico and University of Turin, Italy.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Sustainability* Editorial Office E-mail: sustainability@mdpi.com www.mdpi.com/journal/sustainability

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18